Upload folder using huggingface_hub

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+gradio-demo-img.png filter=lfs diff=lfs merge=lfs -text

.github/workflows/update_space.yml

@@ -0,0 +1,28 @@
+name: Run Python script
+
+on:
+  push:
+    branches:
+      - main
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+
+    steps:
+    - name: Checkout
+      uses: actions/checkout@v2
+
+    - name: Set up Python
+      uses: actions/setup-python@v2
+      with:
+        python-version: '3.9'
+
+    - name: Install Gradio
+      run: python -m pip install gradio
+
+    - name: Log in to Hugging Face
+      run: python -c 'import huggingface_hub; huggingface_hub.login(token="${{ secrets.hf_token }}")'
+
+    - name: Deploy to Spaces
+      run: gradio deploy

README.md

@@ -1,12 +1,7 @@
 ---
-title: CountGD Multi-Modal Open-World Counting
-emoji: 🏃
-colorFrom: blue
-colorTo: purple
+title: CountGD_Multi-Modal_Open-World_Counting
+app_file: app.py
 sdk: gradio
 sdk_version: 4.37.2
-app_file: app.py
-pinned: false
 ---
-
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+nohup python -u app.py &

app.log

@@ -0,0 +1,15 @@
+nohup: ignoring input
+/home/niki/anaconda3/envs/gradio/lib/python3.9/site-packages/torch/functional.py:507: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at ../aten/src/ATen/native/TensorShape.cpp:3549.)
+  return _VF.meshgrid(tensors, **kwargs)  # type: ignore[attr-defined]
+Some weights of BertModel were not initialized from the model checkpoint at checkpoints/bert-base-uncased and are newly initialized: ['pooler.dense.bias', 'pooler.dense.weight']
+You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.
+huggingface/tokenizers: The current process just got forked, after parallelism has already been used. Disabling parallelism to avoid deadlocks...
+To disable this warning, you can either:
+	- Avoid using `tokenizers` before the fork if possible
+	- Explicitly set the environment variable TOKENIZERS_PARALLELISM=(true | false)
+/home/niki/anaconda3/envs/gradio/lib/python3.9/site-packages/transformers/modeling_utils.py:977: FutureWarning: The `device` argument is deprecated and will be removed in v5 of Transformers.
+  warnings.warn(
+/home/niki/anaconda3/envs/gradio/lib/python3.9/site-packages/torch/utils/checkpoint.py:460: UserWarning: torch.utils.checkpoint: please pass in use_reentrant=True or use_reentrant=False explicitly. The default value of use_reentrant will be updated to be False in the future. To maintain current behavior, pass use_reentrant=True. It is recommended that you use use_reentrant=False. Refer to docs for more details on the differences between the two variants.
+  warnings.warn(
+/home/niki/anaconda3/envs/gradio/lib/python3.9/site-packages/torch/utils/checkpoint.py:90: UserWarning: None of the inputs have requires_grad=True. Gradients will be None
+  warnings.warn(

app.py

@@ -0,0 +1,437 @@
+import gradio as gr
+import copy
+import random
+import torch
+import PIL
+from PIL import Image, ImageDraw, ImageFont
+import torchvision.transforms.functional as F
+import numpy as np
+import argparse
+import json
+import plotly.express as px
+import pandas as pd
+from util.slconfig import SLConfig, DictAction
+from util.misc import nested_tensor_from_tensor_list
+import datasets.transforms as T
+import scipy.ndimage as ndimage
+import matplotlib.pyplot as plt
+from gradio_image_prompter import ImagePrompter
+# https://github.com/PhyscalX/gradio-image-prompter/tree/main/backend/gradio_image_prompter/templates/component
+import io
+from enum import Enum
+import os
+os.environ["GRADIO_TEMP_DIR"] = os.path.join(os.getcwd(), "tmp")
+
+class AppSteps(Enum):
+    JUST_TEXT = 1
+    TEXT_AND_EXEMPLARS = 2
+    JUST_EXEMPLARS = 3
+    FULL_APP = 4
+
+CONF_THRESH = 0.23
+
+# MODEL:
+def get_args_parser():
+    """
+    Example eval command:
+
+    >> python main.py --output_dir ./gdino_test -c config/cfg_fsc147_vit_b_test.py --eval --datasets config/datasets_fsc147.json --pretrain_model_path ../checkpoints_and_logs/gdino_train/checkpoint_best_regular.pth --options text_encoder_type=checkpoints/bert-base-uncased --sam_tt_norm --crop
+    """
+    parser = argparse.ArgumentParser("Set transformer detector", add_help=False)
+    parser.add_argument(
+        "--options",
+        nargs="+",
+        action=DictAction,
+        help="override some settings in the used config, the key-value pair "
+        "in xxx=yyy format will be merged into config file.",
+    )
+
+    # dataset parameters
+    parser.add_argument("--remove_difficult", action="store_true")
+    parser.add_argument("--fix_size", action="store_true")
+
+    # training parameters
+    parser.add_argument("--note", default="", help="add some notes to the experiment")
+    parser.add_argument(
+        "--device", default="cuda", help="device to use for training / testing"
+    )
+    parser.add_argument("--resume", default="", help="resume from checkpoint")
+    parser.add_argument(
+        "--pretrain_model_path",
+        help="load from other checkpoint",
+        default="checkpoint_best_regular.pth",
+    )
+    parser.add_argument("--finetune_ignore", type=str, nargs="+")
+    parser.add_argument(
+        "--start_epoch", default=0, type=int, metavar="N", help="start epoch"
+    )
+    parser.add_argument("--eval", action="store_false")
+    parser.add_argument("--num_workers", default=8, type=int)
+    parser.add_argument("--test", action="store_true")
+    parser.add_argument("--debug", action="store_true")
+    parser.add_argument("--find_unused_params", action="store_true")
+    parser.add_argument("--save_results", action="store_true")
+    parser.add_argument("--save_log", action="store_true")
+
+    # distributed training parameters
+    parser.add_argument(
+        "--world_size", default=1, type=int, help="number of distributed processes"
+    )
+    parser.add_argument(
+        "--dist_url", default="env://", help="url used to set up distributed training"
+    )
+    parser.add_argument(
+        "--rank", default=0, type=int, help="number of distributed processes"
+    )
+    parser.add_argument(
+        "--local_rank", type=int, help="local rank for DistributedDataParallel"
+    )
+    parser.add_argument(
+        "--local-rank", type=int, help="local rank for DistributedDataParallel"
+    )
+    parser.add_argument("--amp", action="store_true", help="Train with mixed precision")
+    return parser
+
+# Get counting model.
+def build_model_and_transforms(args):
+    normalize = T.Compose(
+        [T.ToTensor(), T.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])]
+    )
+    data_transform = T.Compose(
+        [
+            T.RandomResize([800], max_size=1333),
+            normalize,
+        ]
+    )
+    cfg = SLConfig.fromfile("cfg_app.py")
+    cfg.merge_from_dict({"text_encoder_type": "checkpoints/bert-base-uncased"})
+    cfg_dict = cfg._cfg_dict.to_dict()
+    args_vars = vars(args)
+    for k, v in cfg_dict.items():
+        if k not in args_vars:
+            setattr(args, k, v)
+        else:
+            raise ValueError("Key {} can used by args only".format(k))
+
+    device = torch.device(args.device)
+    # fix the seed for reproducibility
+    seed = 42
+    torch.manual_seed(seed)
+    np.random.seed(seed)
+    random.seed(seed)
+
+    # we use register to maintain models from catdet6 on.
+    from models.registry import MODULE_BUILD_FUNCS
+
+    assert args.modelname in MODULE_BUILD_FUNCS._module_dict
+
+    build_func = MODULE_BUILD_FUNCS.get(args.modelname)
+    model, _, _ = build_func(args)
+
+    model.to(device)
+
+    checkpoint = torch.load(args.pretrain_model_path, map_location="cpu")["model"]
+    model.load_state_dict(checkpoint, strict=False)
+
+    model.eval()
+
+    return model, data_transform
+
+
+parser = argparse.ArgumentParser("Counting Application", parents=[get_args_parser()])
+args = parser.parse_args()
+model, transform = build_model_and_transforms(args)
+
+examples = [
+    ["strawberry.jpg", "strawberry", {"image": "strawberry.jpg"}],
+    ["strawberry.jpg", "blueberry", {"image": "strawberry.jpg"}],
+    ["bird-1.JPG", "bird", {"image": "bird-2.JPG"}],
+    ["fish.jpg", "fish", {"image": "fish.jpg"}],
+    ["women.jpg", "girl", {"image": "women.jpg"}],
+    ["women.jpg", "boy", {"image": "women.jpg"}],
+    ["balloon.jpg", "hot air balloon", {"image": "balloon.jpg"}],
+    ["deer.jpg", "deer", {"image": "deer.jpg"}],
+    ["apple.jpg", "apple", {"image": "apple.jpg"}],
+    ["egg.jpg", "egg", {"image": "egg.jpg"}],
+    ["stamp.jpg", "stamp", {"image": "stamp.jpg"}],
+    ["green-pea.jpg", "green pea", {"image": "green-pea.jpg"}],
+    ["lego.jpg", "lego", {"image": "lego.jpg"}]
+]
+
+# APP:
+def get_box_inputs(prompts):
+    box_inputs = []
+    for prompt in prompts:
+        if prompt[2] == 2.0 and prompt[5] == 3.0:
+            box_inputs.append([prompt[0], prompt[1], prompt[3], prompt[4]])
+
+    return box_inputs
+
+def get_ind_to_filter(text, word_ids, keywords):
+    if len(keywords) <= 0:
+        return list(range(len(word_ids)))
+    input_words = text.split()
+    keywords = keywords.split(",")
+    keywords = [keyword.strip() for keyword in keywords]
+
+    word_inds = []
+    for keyword in keywords:
+        if keyword in input_words:
+            if len(word_inds) <= 0:
+                ind = input_words.index(keyword)
+                word_inds.append(ind)
+            else:
+                ind = input_words.index(keyword, word_inds[-1])
+                word_inds.append(ind)
+        else:
+            raise Exception("Only specify keywords in the input text!")
+
+    inds_to_filter = []
+    for ind in range(len(word_ids)):
+        word_id = word_ids[ind]
+        if word_id in word_inds:
+            inds_to_filter.append(ind)
+
+    return inds_to_filter
+
+def count(image, text, prompts, state):
+    print("state: " + str(state))
+    keywords = "" # do not handle this for now
+    # Handle no prompt case.
+    if prompts is None:
+        prompts = {"image": image, "points": []}
+    input_image, _ = transform(image, {"exemplars": torch.tensor([])})
+    input_image = input_image.unsqueeze(0).cuda()
+    exemplars = get_box_inputs(prompts["points"])
+    print(exemplars)
+    input_image_exemplars, exemplars = transform(prompts["image"], {"exemplars": torch.tensor(exemplars)})
+    input_image_exemplars = input_image_exemplars.unsqueeze(0).cuda()
+    exemplars = [exemplars["exemplars"].cuda()]
+    
+    with torch.no_grad():
+        model_output = model(
+                nested_tensor_from_tensor_list(input_image),
+                nested_tensor_from_tensor_list(input_image_exemplars),
+                exemplars,
+                [torch.tensor([0]).cuda() for _ in range(len(input_image))],
+                captions=[text + " ."] * len(input_image),
+            )
+    
+    ind_to_filter = get_ind_to_filter(text, model_output["token"][0].word_ids, keywords)
+    print(model_output["token"][0].tokens)
+    print(ind_to_filter)
+    print(model_output["pred_logits"].sigmoid()[0].shape)
+    logits = model_output["pred_logits"].sigmoid()[0][:, ind_to_filter]
+    print(logits.shape)
+    boxes = model_output["pred_boxes"][0]
+    if len(keywords.strip()) > 0:
+        box_mask = (logits > CONF_THRESH).sum(dim=-1) == len(ind_to_filter)
+    else:
+        box_mask = logits.max(dim=-1).values > CONF_THRESH
+    logits = logits[box_mask, :].cpu().numpy()
+    boxes = boxes[box_mask, :].cpu().numpy()
+    
+    # Plot results.
+    (w, h) = image.size
+    det_map = np.zeros((h, w))
+    det_map[(h * boxes[:, 1]).astype(int), (w * boxes[:, 0]).astype(int)] = 1
+    det_map = ndimage.gaussian_filter(
+        det_map, sigma=(w // 200, w // 200), order=0
+    )
+    plt.imshow(image)
+    plt.imshow(det_map[None, :].transpose(1, 2, 0), 'jet', interpolation='none', alpha=0.7)
+    plt.axis('off')
+    img_buf = io.BytesIO()
+    plt.savefig(img_buf, format='png', bbox_inches='tight')
+
+    output_img = Image.open(img_buf)
+
+    if AppSteps.TEXT_AND_EXEMPLARS not in state:
+        exemplar_image = ImagePrompter(type='pil', label='Visual Exemplar Image', value=prompts, interactive=True, visible=True)
+        new_submit_btn = gr.Button("Count", variant="primary", interactive=False)
+        state = [AppSteps.JUST_TEXT, AppSteps.TEXT_AND_EXEMPLARS]
+        main_instructions_comp = gr.Markdown(visible=False)
+        step_3 = gr.Tab(visible=False)
+    elif AppSteps.FULL_APP not in state:
+        exemplar_image = ImagePrompter(type='pil', label='Visual Exemplar Image', value=prompts, interactive=True, visible=True)
+        new_submit_btn = submit_btn
+        state = [AppSteps.JUST_TEXT, AppSteps.TEXT_AND_EXEMPLARS, AppSteps.FULL_APP]
+        main_instructions_comp = gr.Markdown(visible=True)
+        step_3 = gr.Tab(visible=True)
+    else:
+        exemplar_image = ImagePrompter(type='pil', label='Visual Exemplar Image', value=prompts, interactive=True, visible=True)
+        new_submit_btn = submit_btn
+        main_instructions_comp = gr.Markdown(visible=True)
+        step_3 = gr.Tab(visible=True)
+
+    out_label = "Detected instances predicted with"
+    if len(text.strip()) > 0:
+        out_label += " text"
+        if exemplars[0].size()[0] == 1:
+            out_label += " and " + str(exemplars[0].size()[0]) + " visual exemplar."        
+        elif exemplars[0].size()[0] > 1:
+            out_label += " and " + str(exemplars[0].size()[0]) + " visual exemplars."
+        else:
+            out_label += "."
+    elif exemplars[0].size()[0] > 0:
+        if exemplars[0].size()[0] == 1:
+            out_label += " " + str(exemplars[0].size()[0]) + " visual exemplar."
+        else:
+            out_label += " " + str(exemplars[0].size()[0]) + " visual exemplars."
+    else:
+        out_label = "Nothing specified to detect."
+    return (gr.Image(output_img, visible=True, label=out_label, show_label=True), gr.Number(label="Predicted Count", visible=True, value=boxes.shape[0]), new_submit_btn, gr.Tab(visible=True), step_3, state)
+
+def count_main(image, text, prompts):
+    keywords = "" # do not handle this for now
+    # Handle no prompt case.
+    if prompts is None:
+        prompts = {"image": image, "points": []}
+    input_image, _ = transform(image, {"exemplars": torch.tensor([])})
+    input_image = input_image.unsqueeze(0).cuda()
+    exemplars = get_box_inputs(prompts["points"])
+    print(exemplars)
+    input_image_exemplars, exemplars = transform(prompts["image"], {"exemplars": torch.tensor(exemplars)})
+    input_image_exemplars = input_image_exemplars.unsqueeze(0).cuda()
+    exemplars = [exemplars["exemplars"].cuda()]
+    
+    with torch.no_grad():
+        model_output = model(
+                nested_tensor_from_tensor_list(input_image),
+                nested_tensor_from_tensor_list(input_image_exemplars),
+                exemplars,
+                [torch.tensor([0]).cuda() for _ in range(len(input_image))],
+                captions=[text + " ."] * len(input_image),
+            )
+    
+    ind_to_filter = get_ind_to_filter(text, model_output["token"][0].word_ids, keywords)
+    print(model_output["token"][0].tokens)
+    print(ind_to_filter)
+    print(model_output["pred_logits"].sigmoid()[0].shape)
+    logits = model_output["pred_logits"].sigmoid()[0][:, ind_to_filter]
+    print(logits.shape)
+    boxes = model_output["pred_boxes"][0]
+    if len(keywords.strip()) > 0:
+        box_mask = (logits > CONF_THRESH).sum(dim=-1) == len(ind_to_filter)
+    else:
+        box_mask = logits.max(dim=-1).values > CONF_THRESH
+    logits = logits[box_mask, :].cpu().numpy()
+    boxes = boxes[box_mask, :].cpu().numpy()
+    
+    # Plot results.
+    (w, h) = image.size
+    det_map = np.zeros((h, w))
+    det_map[(h * boxes[:, 1]).astype(int), (w * boxes[:, 0]).astype(int)] = 1
+    det_map = ndimage.gaussian_filter(
+        det_map, sigma=(w // 200, w // 200), order=0
+    )
+    plt.imshow(image)
+    plt.imshow(det_map[None, :].transpose(1, 2, 0), 'jet', interpolation='none', alpha=0.7)
+    plt.axis('off')
+    img_buf = io.BytesIO()
+    plt.savefig(img_buf, format='png', bbox_inches='tight')
+
+    output_img = Image.open(img_buf)
+
+    out_label = "Detected instances predicted with"
+    if len(text.strip()) > 0:
+        out_label += " text"
+        if exemplars[0].size()[0] == 1:
+            out_label += " and " + str(exemplars[0].size()[0]) + " visual exemplar."        
+        elif exemplars[0].size()[0] > 1:
+            out_label += " and " + str(exemplars[0].size()[0]) + " visual exemplars."
+        else:
+            out_label += "."
+    elif exemplars[0].size()[0] > 0:
+        if exemplars[0].size()[0] == 1:
+            out_label += " " + str(exemplars[0].size()[0]) + " visual exemplar."
+        else:
+            out_label += " " + str(exemplars[0].size()[0]) + " visual exemplars."
+    else:
+        out_label = "Nothing specified to detect."
+    return (gr.Image(output_img, visible=True, label=out_label, show_label=True), gr.Number(label="Predicted Count", visible=True, value=boxes.shape[0]))
+
+def remove_label(image):
+    return gr.Image(show_label=False)
+
+def check_submit_btn(exemplar_image_prompts, state):
+    if AppSteps.TEXT_AND_EXEMPLARS not in state or len(state) == 3:
+        return gr.Button("Count", variant="primary", interactive=True)
+    elif exemplar_image_prompts is None:
+        return gr.Button("Count", variant="primary", interactive=False)
+    elif len(get_box_inputs(exemplar_image_prompts["points"])) > 0:
+        return gr.Button("Count", variant="primary", interactive=True)
+    else:
+        return gr.Button("Count", variant="primary", interactive=False)
+
+exemplar_img_drawing_instructions_part_1 = '<p><strong>Congrats, you have counted the strawberries!</strong> You can also draw a box around the object you want to count. <strong>Click and drag the mouse on the image below to draw a box around one of the strawberries.</strong> You can click the back button in the top right of the image to delete the box and try again.<img src="file/button-legend.jpg" width="750"></p>'
+exemplar_img_drawing_instructions_part_2 = '<p>The boxes you draw are called \"visual exemplars,\" image examples of what you want the model to count. You can add more boxes around more examples of strawberries in the image above to increase the accuracy of the predicted count. You can also use strawberries from a different image to specify the object to count by uploading or pasting a new image above and drawing boxes around strawberries in it.</p>'
+instructions_main = """
+# How to Use the App
+As shown earlier, there are 3 ways to specify the object to count: (1) with text only, (2) with text and any number of boxes (i.e., "visual exemplars") around example objects, and (3) with visual exemplars only. What is being used is indicated in the top left of the output image. How to try each case is detailed below.
+
+<ol>
+  <li><strong>Text Only: </strong> Only provide text describing the object to count in the textbox titled "What would you like to count?" Delete all boxes drawn on the visual exemplar image.</li>
+  <li><strong>Text + Visual Exemplars: </strong> Provide text describing the object to count in the textbox titled "What would you like to count?" and draw at least one box around an example object in the visual exemplar image.</li>
+  <li><strong>Visual Exemplars Only: </strong> Remove all text in the textbox titled "What would you like to count?" and draw at least one box around an example object in the visual exemplar image.</li>
+</ol>
+
+## Click on the "App" tab at the top of the screen to exit the tutorial and start using the main app!
+"""
+
+with gr.Blocks(title="CountGD: Multi-Modal Open-World Counting", theme="soft", head="""<meta name="viewport" content="width=device-width, initial-scale=1, user-scalable=1">""") as demo:
+    state = gr.State(value=[AppSteps.JUST_TEXT])
+    with gr.Tab("Tutorial"):
+        with gr.Row():
+            with gr.Column():
+                with gr.Tab("Step 3", visible=False) as step_3:
+                    main_instructions = gr.Markdown(instructions_main)
+                with gr.Tab("Step 2", visible=False) as step_2:
+                    gr.Markdown(exemplar_img_drawing_instructions_part_1)
+                    exemplar_image = ImagePrompter(type='pil', label='Visual Exemplar Image', show_label=True, value={"image": "strawberry.jpg", "points": []}, interactive=True)
+                    with gr.Accordion("Open for Further Information", open=False):
+                        gr.Markdown(exemplar_img_drawing_instructions_part_2)
+                with gr.Tab("Step 1", visible=True) as step_1:
+                    input_image = gr.Image(type='pil', label='Input Image', show_label='True', value="strawberry.jpg", interactive=False, width="30vw")
+                    gr.Markdown('# Click "Count" to count the strawberries.')
+                
+            with gr.Column():
+                with gr.Tab("Output Image"):
+                    detected_instances = gr.Image(label="Detected Instances", show_label='True', interactive=False, visible=True, width="40vw")
+                    
+        with gr.Row():
+            input_text = gr.Textbox(label="What would you like to count?", value="strawberry", interactive=True)
+            pred_count = gr.Number(label="Predicted Count", visible=False)
+        submit_btn = gr.Button("Count", variant="primary", interactive=True)
+                
+        submit_btn.click(fn=remove_label, inputs=[detected_instances], outputs=[detected_instances]).then(fn=count, inputs=[input_image, input_text, exemplar_image, state], outputs=[detected_instances, pred_count, submit_btn, step_2, step_3, state])
+        exemplar_image.change(check_submit_btn, inputs=[exemplar_image, state], outputs=[submit_btn])
+    with gr.Tab("App", visible=True) as main_app:
+        
+        gr.Markdown(
+              """
+              # <center>CountGD: Multi-Modal Open-World Counting
+              <center><h3>Count objects with text, visual exemplars, or both together.</h3>
+              <h3>Scroll down to try more examples</h3>
+              <h3><a href='https://github.com/niki-amini-naieni/CountGD/' target='_blank' rel='noopener'>[paper]</a>
+                <a href='https://github.com/niki-amini-naieni/CountGD/' target='_blank' rel='noopener'>[code]</a></h3>
+              Limitation: this app does not support fine-grained counting based on attributes or visual grounding inputs yet.</center>
+              """
+            )
+            
+        with gr.Row():
+            with gr.Column():
+              input_image_main = gr.Image(type='pil', label='Input Image', show_label='True', value="strawberry.jpg", interactive=True)
+              input_text_main = gr.Textbox(label="What would you like to count?", placeholder="", value="strawberry")
+              exemplar_image_main = ImagePrompter(type='pil', label='Visual Exemplar Image', show_label=True, value={"image": "strawberry.jpg", "points": []}, interactive=True)
+            with gr.Column():
+              detected_instances_main = gr.Image(label="Detected Instances", show_label='True', interactive=False)
+              pred_count_main = gr.Number(label="Predicted Count")
+              submit_btn_main = gr.Button("Count", variant="primary")
+              clear_btn_main = gr.ClearButton(variant="secondary")
+        gr.Examples(label="Examples: click on a row to load the example. Add visual exemplars by drawing boxes on the loaded \"Visual Exemplar Image.\"", examples=examples, inputs=[input_image_main, input_text_main, exemplar_image_main])
+        submit_btn_main.click(fn=remove_label, inputs=[detected_instances_main], outputs=[detected_instances_main]).then(fn=count_main, inputs=[input_image_main, input_text_main, exemplar_image_main], outputs=[detected_instances_main, pred_count_main])
+        clear_btn_main.add([input_image_main, input_text_main, exemplar_image_main, detected_instances_main, pred_count_main])
+        
+
+demo.launch(share=True, allowed_paths=['back-icon.jpg', 'paste-icon.jpg', 'upload-icon.jpg', 'button-legend.jpg'])

cfg_app.py

@@ -0,0 +1,118 @@
+data_aug_scales = [480, 512, 544, 576, 608, 640, 672, 704, 736, 768, 800]
+data_aug_max_size = 1333
+data_aug_scales2_resize = [400, 500, 600]
+data_aug_scales2_crop = [384, 600]
+data_aug_scale_overlap = None
+batch_size = 4
+modelname = 'groundingdino'
+backbone = "swin_B_384_22k"
+position_embedding = 'sine'
+pe_temperatureH = 20
+pe_temperatureW = 20
+return_interm_indices = [1, 2, 3]
+enc_layers = 6
+dec_layers = 6
+pre_norm = False
+dim_feedforward = 2048
+hidden_dim = 256
+dropout = 0.0
+nheads = 8
+num_queries = 900
+query_dim = 4
+num_patterns = 0
+num_feature_levels = 4
+enc_n_points = 4
+dec_n_points = 4
+two_stage_type = 'standard'
+two_stage_bbox_embed_share = False
+two_stage_class_embed_share = False
+transformer_activation = 'relu'
+dec_pred_bbox_embed_share = True
+dn_box_noise_scale = 1.0
+dn_label_noise_ratio = 0.5
+dn_label_coef = 1.0
+dn_bbox_coef = 1.0
+embed_init_tgt = True
+dn_labelbook_size = 91
+max_text_len = 256
+text_encoder_type = "bert-base-uncased"
+use_text_enhancer = True
+use_fusion_layer = True
+use_checkpoint = True
+use_transformer_ckpt = True
+use_text_cross_attention = True
+text_dropout = 0.0
+fusion_dropout = 0.0
+fusion_droppath = 0.1
+sub_sentence_present = True
+max_labels = 90                               # pos + neg
+lr = 0.0001                                   # base learning rate
+backbone_freeze_keywords = None               # only for gdino backbone
+freeze_keywords = ['backbone.0', 'bert']      # for whole model, e.g. ['backbone.0', 'bert'] for freeze visual encoder and text encoder
+lr_backbone = 1e-05                           # specific learning rate
+lr_backbone_names = ['backbone.0', 'bert']
+lr_linear_proj_mult = 1e-05
+lr_linear_proj_names = ['ref_point_head', 'sampling_offsets']
+weight_decay = 0.0001
+param_dict_type = 'ddetr_in_mmdet'
+ddetr_lr_param = False
+epochs = 30
+lr_drop = 10
+save_checkpoint_interval = 10
+clip_max_norm = 0.1
+onecyclelr = False
+multi_step_lr = False
+lr_drop_list = [10, 20]
+frozen_weights = None
+dilation = False
+pdetr3_bbox_embed_diff_each_layer = False
+pdetr3_refHW = -1
+random_refpoints_xy = False
+fix_refpoints_hw = -1
+dabdetr_yolo_like_anchor_update = False
+dabdetr_deformable_encoder = False
+dabdetr_deformable_decoder = False
+use_deformable_box_attn = False
+box_attn_type = 'roi_align'
+dec_layer_number = None
+decoder_layer_noise = False
+dln_xy_noise = 0.2
+dln_hw_noise = 0.2
+add_channel_attention = False
+add_pos_value = False
+two_stage_pat_embed = 0
+two_stage_add_query_num = 0
+two_stage_learn_wh = False
+two_stage_default_hw = 0.05
+two_stage_keep_all_tokens = False
+num_select = 900
+batch_norm_type = 'FrozenBatchNorm2d'
+masks = False
+aux_loss = True
+set_cost_class = 5.0
+set_cost_bbox = 1.0
+set_cost_giou = 0.0
+cls_loss_coef = 5.0
+bbox_loss_coef = 1.0
+giou_loss_coef = 0.0
+enc_loss_coef = 1.0
+interm_loss_coef = 1.0
+no_interm_box_loss = False
+mask_loss_coef = 1.0
+dice_loss_coef = 1.0
+focal_alpha = 0.25
+focal_gamma = 2.0
+decoder_sa_type = 'sa'
+matcher_type = 'HungarianMatcher'
+decoder_module_seq = ['sa', 'ca', 'ffn']
+nms_iou_threshold = -1
+dec_pred_class_embed_share = True
+match_unstable_error = True
+use_detached_boxes_dec_out = False
+dn_scalar = 100
+
+box_threshold = 0.23
+text_threshold = 0
+use_coco_eval = False
+label_list = ['alcohol bottle', 'baguette roll', 'ball', 'banana', 'bead', 'bee', 'birthday candle', 'biscuit', 'boat', 'bottle', 'bowl', 'box', 'bread roll', 'brick', 'buffalo', 'bun', 'calamari ring', 'can', 'candle', 'cap', 'car', 'cartridge', 'cassette', 'cement bag', 'cereal', 'chewing gum piece', 'chopstick', 'clam', 'coffee bean', 'coin', 'cotton ball', 'cow', 'crane', 'crayon', 'croissant', 'crow', 'cup', 'cupcake', 'cupcake holder', 'fish', 'gemstone', 'go game piece', 'goat', 'goldfish snack', 'goose', 'ice cream', 'ice cream cone', 'instant noodle', 'jade stone', 'jeans', 'kidney bean', 'kitchen towel', 'lighter', 'lipstick', 'm&m piece', 'macaron', 'match', 'meat skewer', 'mini blind', 'mosaic tile', 'naan bread', 'nail', 'nut', 'onion ring', 'orange', 'pearl', 'pen', 'pencil', 'penguin', 'pepper', 'person', 'pigeon', 'plate', 'polka dot tile', 'potato', 'rice bag', 'roof tile', 'screw', 'shoe', 'spoon', 'spring roll', 'stair', 'stapler pin', 'straw', 'supermarket shelf', 'swan', 'tomato', 'watermelon', 'window', 'zebra']
+val_label_list = ["apple", "candy piece", "carrom board piece", "cashew nut", "comic book", "crab cake", "deer", "egg", "elephant", "finger food", "green pea", "hot air balloon", "keyboard key", "lego", "marble", "marker", "nail polish", "potato chip", "red bean", "round dessert", "sauce bottle", "sea shell", "sheep", "ski", "stamp", "sticky note", "strawberry", "sunglasses", "tree log", "watch"]

checkpoint_best_regular.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c1bab864b17db345b4c6e3aaabb5765bc2c0a90d0bc8defb5e664a74a50aa126
+size 1250122522

checkpoints/bert-base-uncased/config.json

@@ -0,0 +1,26 @@
+{
+  "_name_or_path": "bert-base-uncased",
+  "architectures": [
+    "BertModel"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "classifier_dropout": null,
+  "gradient_checkpointing": false,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "layer_norm_eps": 1e-12,
+  "max_position_embeddings": 512,
+  "model_type": "bert",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "pad_token_id": 0,
+  "position_embedding_type": "absolute",
+  "torch_dtype": "float32",
+  "transformers_version": "4.39.1",
+  "type_vocab_size": 2,
+  "use_cache": true,
+  "vocab_size": 30522
+}

checkpoints/bert-base-uncased/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5875f83030335d194f35b15a32e7f4e654aa302aa83af032a3f36d035dcaf8af
+size 435588776

checkpoints/bert-base-uncased/special_tokens_map.json

@@ -0,0 +1,7 @@
+{
+  "cls_token": "[CLS]",
+  "mask_token": "[MASK]",
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "unk_token": "[UNK]"
+}

checkpoints/bert-base-uncased/tokenizer_config.json

@@ -0,0 +1,55 @@
+{
+  "added_tokens_decoder": {
+    "0": {
+      "content": "[PAD]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "100": {
+      "content": "[UNK]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "101": {
+      "content": "[CLS]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "102": {
+      "content": "[SEP]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "103": {
+      "content": "[MASK]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "clean_up_tokenization_spaces": true,
+  "cls_token": "[CLS]",
+  "do_lower_case": true,
+  "mask_token": "[MASK]",
+  "model_max_length": 512,
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "strip_accents": null,
+  "tokenize_chinese_chars": true,
+  "tokenizer_class": "BertTokenizer",
+  "unk_token": "[UNK]"
+}

checkpoints/groundingdino_swinb_cogcoor.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:46270f7a822e6906b655b729c90613e48929d0f2bb8b9b76fd10a856f3ac6ab7
+size 938057991

datasets/__init__.py

@@ -0,0 +1,23 @@
+# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved
+import torch.utils.data
+import torchvision
+from .coco import build as build_coco
+
+
+def get_coco_api_from_dataset(dataset):
+    for _ in range(10):
+        # if isinstance(dataset, torchvision.datasets.CocoDetection):
+        #     break
+        if isinstance(dataset, torch.utils.data.Subset):
+            dataset = dataset.dataset
+    if isinstance(dataset, torchvision.datasets.CocoDetection):
+        return dataset.coco
+
+
+def build_dataset(image_set, args, datasetinfo):
+    if datasetinfo["dataset_mode"] == 'coco':
+        return build_coco(image_set, args, datasetinfo)
+    if datasetinfo["dataset_mode"] == 'odvg':
+        from .odvg import build_odvg
+        return build_odvg(image_set, args, datasetinfo)
+    raise ValueError(f'dataset {args.dataset_file} not supported')

datasets/coco.py

@@ -0,0 +1,827 @@
+# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved
+"""
+COCO dataset which returns image_id for evaluation.
+
+Mostly copy-paste from https://github.com/pytorch/vision/blob/13b35ff/references/detection/coco_utils.py
+"""
+if __name__ == "__main__":
+    # for debug only
+    import os, sys
+
+    sys.path.append(os.path.dirname(sys.path[0]))
+from torchvision.datasets.vision import VisionDataset
+
+import json
+from pathlib import Path
+import random
+import os
+from typing import Any, Callable, List, Optional, Tuple
+
+from PIL import Image
+
+import torch
+import torch.utils.data
+import torchvision
+from pycocotools import mask as coco_mask
+
+from datasets.data_util import preparing_dataset
+import datasets.transforms as T
+from util.box_ops import box_cxcywh_to_xyxy, box_iou
+
+__all__ = ["build"]
+
+
+class label2compat:
+    def __init__(self) -> None:
+        self.category_map_str = {
+            "1": 1,
+            "2": 2,
+            "3": 3,
+            "4": 4,
+            "5": 5,
+            "6": 6,
+            "7": 7,
+            "8": 8,
+            "9": 9,
+            "10": 10,
+            "11": 11,
+            "13": 12,
+            "14": 13,
+            "15": 14,
+            "16": 15,
+            "17": 16,
+            "18": 17,
+            "19": 18,
+            "20": 19,
+            "21": 20,
+            "22": 21,
+            "23": 22,
+            "24": 23,
+            "25": 24,
+            "27": 25,
+            "28": 26,
+            "31": 27,
+            "32": 28,
+            "33": 29,
+            "34": 30,
+            "35": 31,
+            "36": 32,
+            "37": 33,
+            "38": 34,
+            "39": 35,
+            "40": 36,
+            "41": 37,
+            "42": 38,
+            "43": 39,
+            "44": 40,
+            "46": 41,
+            "47": 42,
+            "48": 43,
+            "49": 44,
+            "50": 45,
+            "51": 46,
+            "52": 47,
+            "53": 48,
+            "54": 49,
+            "55": 50,
+            "56": 51,
+            "57": 52,
+            "58": 53,
+            "59": 54,
+            "60": 55,
+            "61": 56,
+            "62": 57,
+            "63": 58,
+            "64": 59,
+            "65": 60,
+            "67": 61,
+            "70": 62,
+            "72": 63,
+            "73": 64,
+            "74": 65,
+            "75": 66,
+            "76": 67,
+            "77": 68,
+            "78": 69,
+            "79": 70,
+            "80": 71,
+            "81": 72,
+            "82": 73,
+            "84": 74,
+            "85": 75,
+            "86": 76,
+            "87": 77,
+            "88": 78,
+            "89": 79,
+            "90": 80,
+        }
+        self.category_map = {int(k): v for k, v in self.category_map_str.items()}
+
+    def __call__(self, target, img=None):
+        labels = target["labels"]
+        res = torch.zeros(labels.shape, dtype=labels.dtype)
+        for idx, item in enumerate(labels):
+            res[idx] = self.category_map[item.item()] - 1
+        target["label_compat"] = res
+        if img is not None:
+            return target, img
+        else:
+            return target
+
+
+class label_compat2onehot:
+    def __init__(self, num_class=80, num_output_objs=1):
+        self.num_class = num_class
+        self.num_output_objs = num_output_objs
+        if num_output_objs != 1:
+            raise DeprecationWarning(
+                "num_output_objs!=1, which is only used for comparison"
+            )
+
+    def __call__(self, target, img=None):
+        labels = target["label_compat"]
+        place_dict = {k: 0 for k in range(self.num_class)}
+        if self.num_output_objs == 1:
+            res = torch.zeros(self.num_class)
+            for i in labels:
+                itm = i.item()
+                res[itm] = 1.0
+        else:
+            # compat with baseline
+            res = torch.zeros(self.num_class, self.num_output_objs)
+            for i in labels:
+                itm = i.item()
+                res[itm][place_dict[itm]] = 1.0
+                place_dict[itm] += 1
+        target["label_compat_onehot"] = res
+        if img is not None:
+            return target, img
+        else:
+            return target
+
+
+class box_label_catter:
+    def __init__(self):
+        pass
+
+    def __call__(self, target, img=None):
+        labels = target["label_compat"]
+        boxes = target["boxes"]
+        box_label = torch.cat((boxes, labels.unsqueeze(-1)), 1)
+        target["box_label"] = box_label
+        if img is not None:
+            return target, img
+        else:
+            return target
+
+
+class RandomSelectBoxlabels:
+    def __init__(
+        self,
+        num_classes,
+        leave_one_out=False,
+        blank_prob=0.8,
+        prob_first_item=0.0,
+        prob_random_item=0.0,
+        prob_last_item=0.8,
+        prob_stop_sign=0.2,
+    ) -> None:
+        self.num_classes = num_classes
+        self.leave_one_out = leave_one_out
+        self.blank_prob = blank_prob
+
+        self.set_state(
+            prob_first_item, prob_random_item, prob_last_item, prob_stop_sign
+        )
+
+    def get_state(self):
+        return [
+            self.prob_first_item,
+            self.prob_random_item,
+            self.prob_last_item,
+            self.prob_stop_sign,
+        ]
+
+    def set_state(
+        self, prob_first_item, prob_random_item, prob_last_item, prob_stop_sign
+    ):
+        sum_prob = prob_first_item + prob_random_item + prob_last_item + prob_stop_sign
+        assert sum_prob - 1 < 1e-6, (
+            f"Sum up all prob = {sum_prob}. prob_first_item:{prob_first_item}"
+            + f"prob_random_item:{prob_random_item}, prob_last_item:{prob_last_item}"
+            + f"prob_stop_sign:{prob_stop_sign}"
+        )
+
+        self.prob_first_item = prob_first_item
+        self.prob_random_item = prob_random_item
+        self.prob_last_item = prob_last_item
+        self.prob_stop_sign = prob_stop_sign
+
+    def sample_for_pred_first_item(self, box_label: torch.FloatTensor):
+        box_label_known = torch.Tensor(0, 5)
+        box_label_unknown = box_label
+        return box_label_known, box_label_unknown
+
+    def sample_for_pred_random_item(self, box_label: torch.FloatTensor):
+        n_select = int(random.random() * box_label.shape[0])
+        box_label = box_label[torch.randperm(box_label.shape[0])]
+        box_label_known = box_label[:n_select]
+        box_label_unknown = box_label[n_select:]
+        return box_label_known, box_label_unknown
+
+    def sample_for_pred_last_item(self, box_label: torch.FloatTensor):
+        box_label_perm = box_label[torch.randperm(box_label.shape[0])]
+        known_label_list = []
+        box_label_known = []
+        box_label_unknown = []
+        for item in box_label_perm:
+            label_i = item[4].item()
+            if label_i in known_label_list:
+                box_label_known.append(item)
+            else:
+                # first item
+                box_label_unknown.append(item)
+                known_label_list.append(label_i)
+        box_label_known = (
+            torch.stack(box_label_known)
+            if len(box_label_known) > 0
+            else torch.Tensor(0, 5)
+        )
+        box_label_unknown = (
+            torch.stack(box_label_unknown)
+            if len(box_label_unknown) > 0
+            else torch.Tensor(0, 5)
+        )
+        return box_label_known, box_label_unknown
+
+    def sample_for_pred_stop_sign(self, box_label: torch.FloatTensor):
+        box_label_unknown = torch.Tensor(0, 5)
+        box_label_known = box_label
+        return box_label_known, box_label_unknown
+
+    def __call__(self, target, img=None):
+        box_label = target["box_label"]  # K, 5
+
+        dice_number = random.random()
+
+        if dice_number < self.prob_first_item:
+            box_label_known, box_label_unknown = self.sample_for_pred_first_item(
+                box_label
+            )
+        elif dice_number < self.prob_first_item + self.prob_random_item:
+            box_label_known, box_label_unknown = self.sample_for_pred_random_item(
+                box_label
+            )
+        elif (
+            dice_number
+            < self.prob_first_item + self.prob_random_item + self.prob_last_item
+        ):
+            box_label_known, box_label_unknown = self.sample_for_pred_last_item(
+                box_label
+            )
+        else:
+            box_label_known, box_label_unknown = self.sample_for_pred_stop_sign(
+                box_label
+            )
+
+        target["label_onehot_known"] = label2onehot(
+            box_label_known[:, -1], self.num_classes
+        )
+        target["label_onehot_unknown"] = label2onehot(
+            box_label_unknown[:, -1], self.num_classes
+        )
+        target["box_label_known"] = box_label_known
+        target["box_label_unknown"] = box_label_unknown
+
+        return target, img
+
+
+class RandomDrop:
+    def __init__(self, p=0.2) -> None:
+        self.p = p
+
+    def __call__(self, target, img=None):
+        known_box = target["box_label_known"]
+        num_known_box = known_box.size(0)
+        idxs = torch.rand(num_known_box)
+        # indices = torch.randperm(num_known_box)[:int((1-self).p*num_known_box + 0.5 + random.random())]
+        target["box_label_known"] = known_box[idxs > self.p]
+        return target, img
+
+
+class BboxPertuber:
+    def __init__(self, max_ratio=0.02, generate_samples=1000) -> None:
+        self.max_ratio = max_ratio
+        self.generate_samples = generate_samples
+        self.samples = self.generate_pertube_samples()
+        self.idx = 0
+
+    def generate_pertube_samples(self):
+        import torch
+
+        samples = (torch.rand(self.generate_samples, 5) - 0.5) * 2 * self.max_ratio
+        return samples
+
+    def __call__(self, target, img):
+        known_box = target["box_label_known"]  # Tensor(K,5), K known bbox
+        K = known_box.shape[0]
+        known_box_pertube = torch.zeros(K, 6)  # 4:bbox, 1:prob, 1:label
+        if K == 0:
+            pass
+        else:
+            if self.idx + K > self.generate_samples:
+                self.idx = 0
+            delta = self.samples[self.idx : self.idx + K, :]
+            known_box_pertube[:, :4] = known_box[:, :4] + delta[:, :4]
+            iou = (
+                torch.diag(
+                    box_iou(
+                        box_cxcywh_to_xyxy(known_box[:, :4]),
+                        box_cxcywh_to_xyxy(known_box_pertube[:, :4]),
+                    )[0]
+                )
+            ) * (1 + delta[:, -1])
+            known_box_pertube[:, 4].copy_(iou)
+            known_box_pertube[:, -1].copy_(known_box[:, -1])
+
+        target["box_label_known_pertube"] = known_box_pertube
+        return target, img
+
+
+class RandomCutout:
+    def __init__(self, factor=0.5) -> None:
+        self.factor = factor
+
+    def __call__(self, target, img=None):
+        unknown_box = target["box_label_unknown"]  # Ku, 5
+        known_box = target["box_label_known_pertube"]  # Kk, 6
+        Ku = unknown_box.size(0)
+
+        known_box_add = torch.zeros(Ku, 6)  # Ku, 6
+        known_box_add[:, :5] = unknown_box
+        known_box_add[:, 5].uniform_(0.5, 1)
+
+        known_box_add[:, :2] += known_box_add[:, 2:4] * (torch.rand(Ku, 2) - 0.5) / 2
+        known_box_add[:, 2:4] /= 2
+
+        target["box_label_known_pertube"] = torch.cat((known_box, known_box_add))
+        return target, img
+
+
+class RandomSelectBoxes:
+    def __init__(self, num_class=80) -> None:
+        Warning("This is such a slow function and will be deprecated soon!!!")
+        self.num_class = num_class
+
+    def __call__(self, target, img=None):
+        boxes = target["boxes"]
+        labels = target["label_compat"]
+
+        # transform to list of tensors
+        boxs_list = [[] for i in range(self.num_class)]
+        for idx, item in enumerate(boxes):
+            label = labels[idx].item()
+            boxs_list[label].append(item)
+        boxs_list_tensor = [
+            torch.stack(i) if len(i) > 0 else torch.Tensor(0, 4) for i in boxs_list
+        ]
+
+        # random selection
+        box_known = []
+        box_unknown = []
+        for idx, item in enumerate(boxs_list_tensor):
+            ncnt = item.shape[0]
+            nselect = int(
+                random.random() * ncnt
+            )  # close in both sides, much faster than random.randint
+
+            item = item[torch.randperm(ncnt)]
+            # random.shuffle(item)
+            box_known.append(item[:nselect])
+            box_unknown.append(item[nselect:])
+
+        # box_known_tensor = [torch.stack(i) if len(i) > 0 else torch.Tensor(0,4) for i in box_known]
+        # box_unknown_tensor = [torch.stack(i) if len(i) > 0 else torch.Tensor(0,4) for i in box_unknown]
+        # print('box_unknown_tensor:', box_unknown_tensor)
+        target["known_box"] = box_known
+        target["unknown_box"] = box_unknown
+        return target, img
+
+
+def label2onehot(label, num_classes):
+    """
+    label: Tensor(K)
+    """
+    res = torch.zeros(num_classes)
+    for i in label:
+        itm = int(i.item())
+        res[itm] = 1.0
+    return res
+
+
+class MaskCrop:
+    def __init__(self) -> None:
+        pass
+
+    def __call__(self, target, img):
+        known_box = target["known_box"]
+        h, w = img.shape[1:]  # h,w
+        # imgsize = target['orig_size'] # h,w
+
+        scale = torch.Tensor([w, h, w, h])
+
+        # _cnt = 0
+        for boxes in known_box:
+            if boxes.shape[0] == 0:
+                continue
+            box_xyxy = box_cxcywh_to_xyxy(boxes) * scale
+            for box in box_xyxy:
+                x1, y1, x2, y2 = [int(i) for i in box.tolist()]
+                img[:, y1:y2, x1:x2] = 0
+                # _cnt += 1
+        # print("_cnt:", _cnt)
+        return target, img
+
+
+dataset_hook_register = {
+    "label2compat": label2compat,
+    "label_compat2onehot": label_compat2onehot,
+    "box_label_catter": box_label_catter,
+    "RandomSelectBoxlabels": RandomSelectBoxlabels,
+    "RandomSelectBoxes": RandomSelectBoxes,
+    "MaskCrop": MaskCrop,
+    "BboxPertuber": BboxPertuber,
+}
+
+
+class CocoDetection(torchvision.datasets.CocoDetection):
+    def __init__(
+        self, img_folder, ann_file, transforms, return_masks, aux_target_hacks=None
+    ):
+        super(CocoDetection, self).__init__(img_folder, ann_file)
+        self._transforms = transforms
+        self.prepare = ConvertCocoPolysToMask(return_masks)
+        self.aux_target_hacks = aux_target_hacks
+
+    def change_hack_attr(self, hackclassname, attrkv_dict):
+        target_class = dataset_hook_register[hackclassname]
+        for item in self.aux_target_hacks:
+            if isinstance(item, target_class):
+                for k, v in attrkv_dict.items():
+                    setattr(item, k, v)
+
+    def get_hack(self, hackclassname):
+        target_class = dataset_hook_register[hackclassname]
+        for item in self.aux_target_hacks:
+            if isinstance(item, target_class):
+                return item
+
+    def _load_image(self, id: int) -> Image.Image:
+        path = self.coco.loadImgs(id)[0]["file_name"]
+        abs_path = os.path.join(self.root, path)
+        return Image.open(abs_path).convert("RGB")
+
+    def __getitem__(self, idx):
+        """
+        Output:
+            - target: dict of multiple items
+                - boxes: Tensor[num_box, 4]. \
+                    Init type: x0,y0,x1,y1. unnormalized data.
+                    Final type: cx,cy,w,h. normalized data. 
+        """
+        try:
+            img, target = super(CocoDetection, self).__getitem__(idx)
+
+        except:
+            print("Error idx: {}".format(idx))
+            idx += 1
+            img, target = super(CocoDetection, self).__getitem__(idx)
+
+        image_id = self.ids[idx]
+        target = {"image_id": image_id, "annotations": target}
+        exemp_count = 0
+        for instance in target["annotations"]:
+            if instance["area"] != 4:
+                exemp_count += 1
+        # Only provide at most 3 visual exemplars during inference.
+        assert exemp_count == 3
+        img, target = self.prepare(img, target)
+        target["exemplars"] = target["boxes"][-3:]
+        # Remove inaccurate exemplars.
+        if image_id == 6003:
+            target["exemplars"] = torch.tensor([])
+        target["boxes"] = target["boxes"][:-3]
+        target["labels"] = target["labels"][:-3]
+        target["labels_uncropped"] = torch.clone(target["labels"])
+
+        if self._transforms is not None:
+            img, target = self._transforms(img, target)
+
+        # convert to needed format
+        if self.aux_target_hacks is not None:
+            for hack_runner in self.aux_target_hacks:
+                target, img = hack_runner(target, img=img)
+
+        return img, target
+
+
+def convert_coco_poly_to_mask(segmentations, height, width):
+    masks = []
+    for polygons in segmentations:
+        rles = coco_mask.frPyObjects(polygons, height, width)
+        mask = coco_mask.decode(rles)
+        if len(mask.shape) < 3:
+            mask = mask[..., None]
+        mask = torch.as_tensor(mask, dtype=torch.uint8)
+        mask = mask.any(dim=2)
+        masks.append(mask)
+    if masks:
+        masks = torch.stack(masks, dim=0)
+    else:
+        masks = torch.zeros((0, height, width), dtype=torch.uint8)
+    return masks
+
+
+class ConvertCocoPolysToMask(object):
+    def __init__(self, return_masks=False):
+        self.return_masks = return_masks
+
+    def __call__(self, image, target):
+        w, h = image.size
+
+        image_id = target["image_id"]
+        image_id = torch.tensor([image_id])
+
+        anno = target["annotations"]
+
+        anno = [obj for obj in anno if "iscrowd" not in obj or obj["iscrowd"] == 0]
+
+        boxes = [obj["bbox"] for obj in anno]
+        # guard against no boxes via resizing
+        boxes = torch.as_tensor(boxes, dtype=torch.float32).reshape(-1, 4)
+        boxes[:, 2:] += boxes[:, :2]
+        boxes[:, 0::2].clamp_(min=0, max=w)
+        boxes[:, 1::2].clamp_(min=0, max=h)
+
+        classes = [obj["category_id"] for obj in anno]
+        classes = torch.tensor(classes, dtype=torch.int64)
+
+        if self.return_masks:
+            segmentations = [obj["segmentation"] for obj in anno]
+            masks = convert_coco_poly_to_mask(segmentations, h, w)
+
+        keypoints = None
+        if anno and "keypoints" in anno[0]:
+            keypoints = [obj["keypoints"] for obj in anno]
+            keypoints = torch.as_tensor(keypoints, dtype=torch.float32)
+            num_keypoints = keypoints.shape[0]
+            if num_keypoints:
+                keypoints = keypoints.view(num_keypoints, -1, 3)
+
+        keep = (boxes[:, 3] > boxes[:, 1]) & (boxes[:, 2] > boxes[:, 0])
+        boxes = boxes[keep]
+        classes = classes[keep]
+        if self.return_masks:
+            masks = masks[keep]
+        if keypoints is not None:
+            keypoints = keypoints[keep]
+
+        target = {}
+        target["boxes"] = boxes
+        target["labels"] = classes
+        if self.return_masks:
+            target["masks"] = masks
+        target["image_id"] = image_id
+        if keypoints is not None:
+            target["keypoints"] = keypoints
+
+        # for conversion to coco api
+        area = torch.tensor([obj["area"] for obj in anno])
+        iscrowd = torch.tensor(
+            [obj["iscrowd"] if "iscrowd" in obj else 0 for obj in anno]
+        )
+        target["area"] = area[keep]
+        target["iscrowd"] = iscrowd[keep]
+
+        target["orig_size"] = torch.as_tensor([int(h), int(w)])
+        target["size"] = torch.as_tensor([int(h), int(w)])
+
+        return image, target
+
+
+def make_coco_transforms(image_set, fix_size=False, strong_aug=False, args=None):
+    normalize = T.Compose(
+        [T.ToTensor(), T.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])]
+    )
+
+    # config the params for data aug
+    scales = [480, 512, 544, 576, 608, 640, 672, 704, 736, 768, 800]
+    max_size = 1333
+    scales2_resize = [400, 500, 600]
+    scales2_crop = [384, 600]
+
+    # update args from config files
+    scales = getattr(args, "data_aug_scales", scales)
+    max_size = getattr(args, "data_aug_max_size", max_size)
+    scales2_resize = getattr(args, "data_aug_scales2_resize", scales2_resize)
+    scales2_crop = getattr(args, "data_aug_scales2_crop", scales2_crop)
+
+    # resize them
+    data_aug_scale_overlap = getattr(args, "data_aug_scale_overlap", None)
+    if data_aug_scale_overlap is not None and data_aug_scale_overlap > 0:
+        data_aug_scale_overlap = float(data_aug_scale_overlap)
+        scales = [int(i * data_aug_scale_overlap) for i in scales]
+        max_size = int(max_size * data_aug_scale_overlap)
+        scales2_resize = [int(i * data_aug_scale_overlap) for i in scales2_resize]
+        scales2_crop = [int(i * data_aug_scale_overlap) for i in scales2_crop]
+
+    datadict_for_print = {
+        "scales": scales,
+        "max_size": max_size,
+        "scales2_resize": scales2_resize,
+        "scales2_crop": scales2_crop,
+    }
+    # print("data_aug_params:", json.dumps(datadict_for_print, indent=2))
+
+    if image_set == "train":
+        if fix_size:
+            return T.Compose(
+                [
+                    T.RandomHorizontalFlip(),
+                    T.RandomResize([(max_size, max(scales))]),
+                    # T.RandomResize([(512, 512)]),
+                    normalize,
+                ]
+            )
+
+        if strong_aug:
+            import datasets.sltransform as SLT
+
+            return T.Compose(
+                [
+                    T.RandomHorizontalFlip(),
+                    T.RandomSelect(
+                        T.RandomResize(scales, max_size=max_size),
+                        T.Compose(
+                            [
+                                T.RandomResize(scales2_resize),
+                                T.RandomSizeCrop(*scales2_crop),
+                                T.RandomResize(scales, max_size=max_size),
+                            ]
+                        ),
+                    ),
+                    SLT.RandomSelectMulti(
+                        [
+                            SLT.RandomCrop(),
+                            SLT.LightingNoise(),
+                            SLT.AdjustBrightness(2),
+                            SLT.AdjustContrast(2),
+                        ]
+                    ),
+                    normalize,
+                ]
+            )
+
+        return T.Compose(
+            [
+                T.RandomHorizontalFlip(),
+                T.RandomSelect(
+                    T.RandomResize(scales, max_size=max_size),
+                    T.Compose(
+                        [
+                            T.RandomResize(scales2_resize),
+                            T.RandomSizeCrop(*scales2_crop),
+                            T.RandomResize(scales, max_size=max_size),
+                        ]
+                    ),
+                ),
+                normalize,
+            ]
+        )
+
+    if image_set in ["val", "eval_debug", "train_reg", "test"]:
+        if os.environ.get("GFLOPS_DEBUG_SHILONG", False) == "INFO":
+            print("Under debug mode for flops calculation only!!!!!!!!!!!!!!!!")
+            return T.Compose(
+                [
+                    T.ResizeDebug((1280, 800)),
+                    normalize,
+                ]
+            )
+
+        print("max(scales): " + str(max(scales)))
+
+        return T.Compose(
+            [
+                T.RandomResize([max(scales)], max_size=max_size),
+                normalize,
+            ]
+        )
+
+    raise ValueError(f"unknown {image_set}")
+
+
+def get_aux_target_hacks_list(image_set, args):
+    if args.modelname in ["q2bs_mask", "q2bs"]:
+        aux_target_hacks_list = [
+            label2compat(),
+            label_compat2onehot(),
+            RandomSelectBoxes(num_class=args.num_classes),
+        ]
+        if args.masked_data and image_set == "train":
+            # aux_target_hacks_list.append()
+            aux_target_hacks_list.append(MaskCrop())
+    elif args.modelname in [
+        "q2bm_v2",
+        "q2bs_ce",
+        "q2op",
+        "q2ofocal",
+        "q2opclip",
+        "q2ocqonly",
+    ]:
+        aux_target_hacks_list = [
+            label2compat(),
+            label_compat2onehot(),
+            box_label_catter(),
+            RandomSelectBoxlabels(
+                num_classes=args.num_classes,
+                prob_first_item=args.prob_first_item,
+                prob_random_item=args.prob_random_item,
+                prob_last_item=args.prob_last_item,
+                prob_stop_sign=args.prob_stop_sign,
+            ),
+            BboxPertuber(max_ratio=0.02, generate_samples=1000),
+        ]
+    elif args.modelname in ["q2omask", "q2osa"]:
+        if args.coco_aug:
+            aux_target_hacks_list = [
+                label2compat(),
+                label_compat2onehot(),
+                box_label_catter(),
+                RandomSelectBoxlabels(
+                    num_classes=args.num_classes,
+                    prob_first_item=args.prob_first_item,
+                    prob_random_item=args.prob_random_item,
+                    prob_last_item=args.prob_last_item,
+                    prob_stop_sign=args.prob_stop_sign,
+                ),
+                RandomDrop(p=0.2),
+                BboxPertuber(max_ratio=0.02, generate_samples=1000),
+                RandomCutout(factor=0.5),
+            ]
+        else:
+            aux_target_hacks_list = [
+                label2compat(),
+                label_compat2onehot(),
+                box_label_catter(),
+                RandomSelectBoxlabels(
+                    num_classes=args.num_classes,
+                    prob_first_item=args.prob_first_item,
+                    prob_random_item=args.prob_random_item,
+                    prob_last_item=args.prob_last_item,
+                    prob_stop_sign=args.prob_stop_sign,
+                ),
+                BboxPertuber(max_ratio=0.02, generate_samples=1000),
+            ]
+    else:
+        aux_target_hacks_list = None
+
+    return aux_target_hacks_list
+
+
+def build(image_set, args, datasetinfo):
+    img_folder = datasetinfo["root"]
+    ann_file = datasetinfo["anno"]
+
+    # copy to local path
+    if os.environ.get("DATA_COPY_SHILONG") == "INFO":
+        preparing_dataset(
+            dict(img_folder=img_folder, ann_file=ann_file), image_set, args
+        )
+
+    try:
+        strong_aug = args.strong_aug
+    except:
+        strong_aug = False
+    print(img_folder, ann_file)
+    dataset = CocoDetection(
+        img_folder,
+        ann_file,
+        transforms=make_coco_transforms(
+            image_set, fix_size=args.fix_size, strong_aug=strong_aug, args=args
+        ),
+        return_masks=args.masks,
+        aux_target_hacks=None,
+    )
+    return dataset
+
+
+if __name__ == "__main__":
+    # Objects365 Val example
+    dataset_o365 = CocoDetection(
+        "/path/Objects365/train/",
+        "/path/Objects365/slannos/anno_preprocess_train_v2.json",
+        transforms=None,
+        return_masks=False,
+    )
+    print("len(dataset_o365):", len(dataset_o365))

datasets/coco_eval.py

@@ -0,0 +1,266 @@
+# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved
+"""
+COCO evaluator that works in distributed mode.
+
+Mostly copy-paste from https://github.com/pytorch/vision/blob/edfd5a7/references/detection/coco_eval.py
+The difference is that there is less copy-pasting from pycocotools
+in the end of the file, as python3 can suppress prints with contextlib
+"""
+import os
+import contextlib
+import copy
+import numpy as np
+import torch
+
+from pycocotools.cocoeval import COCOeval
+from pycocotools.coco import COCO
+import pycocotools.mask as mask_util
+
+from util.misc import all_gather
+
+
+class CocoEvaluator(object):
+    def __init__(self, coco_gt, iou_types, useCats=True):
+        assert isinstance(iou_types, (list, tuple))
+        coco_gt = copy.deepcopy(coco_gt)
+        self.coco_gt = coco_gt
+
+        self.iou_types = iou_types
+        self.coco_eval = {}
+        for iou_type in iou_types:
+            self.coco_eval[iou_type] = COCOeval(coco_gt, iouType=iou_type)
+            self.coco_eval[iou_type].useCats = useCats
+
+        self.img_ids = []
+        self.eval_imgs = {k: [] for k in iou_types}
+        self.useCats = useCats
+
+    def update(self, predictions):
+        img_ids = list(np.unique(list(predictions.keys())))
+        self.img_ids.extend(img_ids)
+
+        for iou_type in self.iou_types:
+            results = self.prepare(predictions, iou_type)
+
+            # suppress pycocotools prints
+            with open(os.devnull, 'w') as devnull:
+                with contextlib.redirect_stdout(devnull):
+                    coco_dt = COCO.loadRes(self.coco_gt, results) if results else COCO()
+            coco_eval = self.coco_eval[iou_type]
+
+            coco_eval.cocoDt = coco_dt
+            coco_eval.params.imgIds = list(img_ids)
+            coco_eval.params.useCats = self.useCats
+            img_ids, eval_imgs = evaluate(coco_eval)
+
+            self.eval_imgs[iou_type].append(eval_imgs)
+
+    def synchronize_between_processes(self):
+        for iou_type in self.iou_types:
+            self.eval_imgs[iou_type] = np.concatenate(self.eval_imgs[iou_type], 2)
+            create_common_coco_eval(self.coco_eval[iou_type], self.img_ids, self.eval_imgs[iou_type])
+
+    def accumulate(self):
+        for coco_eval in self.coco_eval.values():
+            coco_eval.accumulate()
+
+    def summarize(self):
+        for iou_type, coco_eval in self.coco_eval.items():
+            print("IoU metric: {}".format(iou_type))
+            coco_eval.summarize()
+
+    def prepare(self, predictions, iou_type):
+        if iou_type == "bbox":
+            return self.prepare_for_coco_detection(predictions)
+        elif iou_type == "segm":
+            return self.prepare_for_coco_segmentation(predictions)
+        elif iou_type == "keypoints":
+            return self.prepare_for_coco_keypoint(predictions)
+        else:
+            raise ValueError("Unknown iou type {}".format(iou_type))
+
+    def prepare_for_coco_detection(self, predictions):
+        coco_results = []
+        for original_id, prediction in predictions.items():
+            if len(prediction) == 0:
+                continue
+
+            boxes = prediction["boxes"]
+            boxes = convert_to_xywh(boxes).tolist()
+            if not isinstance(prediction["scores"], list):
+                scores = prediction["scores"].tolist()
+            else:
+                scores = prediction["scores"]
+            if not isinstance(prediction["labels"], list):
+                labels = prediction["labels"].tolist()
+            else:
+                labels = prediction["labels"]
+
+        
+            try:
+                coco_results.extend(
+                    [
+                        {
+                            "image_id": original_id,
+                            "category_id": labels[k],
+                            "bbox": box,
+                            "score": scores[k],
+                        }
+                        for k, box in enumerate(boxes)
+                    ]
+                )
+            except:
+                import ipdb; ipdb.set_trace()
+        return coco_results
+
+    def prepare_for_coco_segmentation(self, predictions):
+        coco_results = []
+        for original_id, prediction in predictions.items():
+            if len(prediction) == 0:
+                continue
+
+            scores = prediction["scores"]
+            labels = prediction["labels"]
+            masks = prediction["masks"]
+
+            masks = masks > 0.5
+
+            scores = prediction["scores"].tolist()
+            labels = prediction["labels"].tolist()
+
+            rles = [
+                mask_util.encode(np.array(mask[0, :, :, np.newaxis], dtype=np.uint8, order="F"))[0]
+                for mask in masks
+            ]
+            for rle in rles:
+                rle["counts"] = rle["counts"].decode("utf-8")
+
+            coco_results.extend(
+                [
+                    {
+                        "image_id": original_id,
+                        "category_id": labels[k],
+                        "segmentation": rle,
+                        "score": scores[k],
+                    }
+                    for k, rle in enumerate(rles)
+                ]
+            )
+        return coco_results
+
+    def prepare_for_coco_keypoint(self, predictions):
+        coco_results = []
+        for original_id, prediction in predictions.items():
+            if len(prediction) == 0:
+                continue
+
+            boxes = prediction["boxes"]
+            boxes = convert_to_xywh(boxes).tolist()
+            scores = prediction["scores"].tolist()
+            labels = prediction["labels"].tolist()
+            keypoints = prediction["keypoints"]
+            keypoints = keypoints.flatten(start_dim=1).tolist()
+
+            coco_results.extend(
+                [
+                    {
+                        "image_id": original_id,
+                        "category_id": labels[k],
+                        'keypoints': keypoint,
+                        "score": scores[k],
+                    }
+                    for k, keypoint in enumerate(keypoints)
+                ]
+            )
+        return coco_results
+
+
+def convert_to_xywh(boxes):
+    xmin, ymin, xmax, ymax = boxes.unbind(1)
+    return torch.stack((xmin, ymin, xmax - xmin, ymax - ymin), dim=1)
+
+
+def merge(img_ids, eval_imgs):
+    all_img_ids = all_gather(img_ids)
+    all_eval_imgs = all_gather(eval_imgs)
+
+    merged_img_ids = []
+    for p in all_img_ids:
+        merged_img_ids.extend(p)
+
+    merged_eval_imgs = []
+    for p in all_eval_imgs:
+        merged_eval_imgs.append(p)
+
+    merged_img_ids = np.array(merged_img_ids)
+    merged_eval_imgs = np.concatenate(merged_eval_imgs, 2)
+
+    # keep only unique (and in sorted order) images
+    merged_img_ids, idx = np.unique(merged_img_ids, return_index=True)
+    merged_eval_imgs = merged_eval_imgs[..., idx]
+
+    return merged_img_ids, merged_eval_imgs
+
+
+def create_common_coco_eval(coco_eval, img_ids, eval_imgs):
+    img_ids, eval_imgs = merge(img_ids, eval_imgs)
+    img_ids = list(img_ids)
+    eval_imgs = list(eval_imgs.flatten())
+
+    coco_eval.evalImgs = eval_imgs
+    coco_eval.params.imgIds = img_ids
+    coco_eval._paramsEval = copy.deepcopy(coco_eval.params)
+
+
+#################################################################
+# From pycocotools, just removed the prints and fixed
+# a Python3 bug about unicode not defined
+#################################################################
+
+
+def evaluate(self):
+    '''
+    Run per image evaluation on given images and store results (a list of dict) in self.evalImgs
+    :return: None
+    '''
+    p = self.params
+    # add backward compatibility if useSegm is specified in params
+    if p.useSegm is not None:
+        p.iouType = 'segm' if p.useSegm == 1 else 'bbox'
+        print('useSegm (deprecated) is not None. Running {} evaluation'.format(p.iouType))
+    p.imgIds = list(np.unique(p.imgIds))
+    if p.useCats:
+        p.catIds = list(np.unique(p.catIds))
+    p.maxDets = sorted(p.maxDets)
+    self.params = p
+
+    self._prepare()
+    # loop through images, area range, max detection number
+    catIds = p.catIds if p.useCats else [-1]
+
+    if p.iouType == 'segm' or p.iouType == 'bbox':
+        computeIoU = self.computeIoU
+    elif p.iouType == 'keypoints':
+        computeIoU = self.computeOks
+    self.ious = {
+        (imgId, catId): computeIoU(imgId, catId)
+        for imgId in p.imgIds
+        for catId in catIds}
+
+    evaluateImg = self.evaluateImg
+    maxDet = p.maxDets[-1]
+    evalImgs = [
+        evaluateImg(imgId, catId, areaRng, maxDet)
+        for catId in catIds
+        for areaRng in p.areaRng
+        for imgId in p.imgIds
+    ]
+    # this is NOT in the pycocotools code, but could be done outside
+    evalImgs = np.asarray(evalImgs).reshape(len(catIds), len(p.areaRng), len(p.imgIds))
+    self._paramsEval = copy.deepcopy(self.params)
+
+    return p.imgIds, evalImgs
+
+#################################################################
+# end of straight copy from pycocotools, just removing the prints
+#################################################################

datasets/coco_panoptic.py

@@ -0,0 +1,99 @@
+# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved
+import json
+from pathlib import Path
+
+import numpy as np
+import torch
+from PIL import Image
+
+from panopticapi.utils import rgb2id
+from util.box_ops import masks_to_boxes
+
+from .coco import make_coco_transforms
+
+
+class CocoPanoptic:
+    def __init__(self, img_folder, ann_folder, ann_file, transforms=None, return_masks=True):
+        with open(ann_file, 'r') as f:
+            self.coco = json.load(f)
+
+        # sort 'images' field so that they are aligned with 'annotations'
+        # i.e., in alphabetical order
+        self.coco['images'] = sorted(self.coco['images'], key=lambda x: x['id'])
+        # sanity check
+        if "annotations" in self.coco:
+            for img, ann in zip(self.coco['images'], self.coco['annotations']):
+                assert img['file_name'][:-4] == ann['file_name'][:-4]
+
+        self.img_folder = img_folder
+        self.ann_folder = ann_folder
+        self.ann_file = ann_file
+        self.transforms = transforms
+        self.return_masks = return_masks
+
+    def __getitem__(self, idx):
+        ann_info = self.coco['annotations'][idx] if "annotations" in self.coco else self.coco['images'][idx]
+        img_path = Path(self.img_folder) / ann_info['file_name'].replace('.png', '.jpg')
+        ann_path = Path(self.ann_folder) / ann_info['file_name']
+
+        img = Image.open(img_path).convert('RGB')
+        w, h = img.size
+        if "segments_info" in ann_info:
+            masks = np.asarray(Image.open(ann_path), dtype=np.uint32)
+            masks = rgb2id(masks)
+
+            ids = np.array([ann['id'] for ann in ann_info['segments_info']])
+            masks = masks == ids[:, None, None]
+
+            masks = torch.as_tensor(masks, dtype=torch.uint8)
+            labels = torch.tensor([ann['category_id'] for ann in ann_info['segments_info']], dtype=torch.int64)
+
+        target = {}
+        target['image_id'] = torch.tensor([ann_info['image_id'] if "image_id" in ann_info else ann_info["id"]])
+        if self.return_masks:
+            target['masks'] = masks
+        target['labels'] = labels
+
+        target["boxes"] = masks_to_boxes(masks)
+
+        target['size'] = torch.as_tensor([int(h), int(w)])
+        target['orig_size'] = torch.as_tensor([int(h), int(w)])
+        if "segments_info" in ann_info:
+            for name in ['iscrowd', 'area']:
+                target[name] = torch.tensor([ann[name] for ann in ann_info['segments_info']])
+
+        if self.transforms is not None:
+            img, target = self.transforms(img, target)
+
+        return img, target
+
+    def __len__(self):
+        return len(self.coco['images'])
+
+    def get_height_and_width(self, idx):
+        img_info = self.coco['images'][idx]
+        height = img_info['height']
+        width = img_info['width']
+        return height, width
+
+
+def build(image_set, args):
+    img_folder_root = Path(args.coco_path)
+    ann_folder_root = Path(args.coco_panoptic_path)
+    assert img_folder_root.exists(), f'provided COCO path {img_folder_root} does not exist'
+    assert ann_folder_root.exists(), f'provided COCO path {ann_folder_root} does not exist'
+    mode = 'panoptic'
+    PATHS = {
+        "train": ("train2017", Path("annotations") / f'{mode}_train2017.json'),
+        "val": ("val2017", Path("annotations") / f'{mode}_val2017.json'),
+    }
+
+    img_folder, ann_file = PATHS[image_set]
+    img_folder_path = img_folder_root / img_folder
+    ann_folder = ann_folder_root / f'{mode}_{img_folder}'
+    ann_file = ann_folder_root / ann_file
+
+    dataset = CocoPanoptic(img_folder_path, ann_folder, ann_file,
+                           transforms=make_coco_transforms(image_set), return_masks=args.masks)
+
+    return dataset

datasets/cocogrounding_eval.py

@@ -0,0 +1,271 @@
+# ------------------------------------------------------------------------
+# Grounding DINO. Midified by Shilong Liu.
+# url: https://github.com/IDEA-Research/GroundingDINO
+# Copyright (c) 2023 IDEA. All Rights Reserved.
+# Licensed under the Apache License, Version 2.0 [see LICENSE for details]
+# ------------------------------------------------------------------------
+# Copyright (c) Aishwarya Kamath & Nicolas Carion. Licensed under the Apache License 2.0. All Rights Reserved
+# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved
+"""
+COCO evaluator that works in distributed mode.
+
+Mostly copy-paste from https://github.com/pytorch/vision/blob/edfd5a7/references/detection/coco_eval.py
+The difference is that there is less copy-pasting from pycocotools
+in the end of the file, as python3 can suppress prints with contextlib
+"""
+import contextlib
+import copy
+import os
+
+import numpy as np
+import pycocotools.mask as mask_util
+import torch
+from pycocotools.coco import COCO
+from pycocotools.cocoeval import COCOeval
+
+from groundingdino.util.misc import all_gather
+
+
+class CocoGroundingEvaluator(object):
+    def __init__(self, coco_gt, iou_types, useCats=True):
+        assert isinstance(iou_types, (list, tuple))
+        coco_gt = copy.deepcopy(coco_gt)
+        self.coco_gt = coco_gt
+
+        self.iou_types = iou_types
+        self.coco_eval = {}
+        for iou_type in iou_types:
+            self.coco_eval[iou_type] = COCOeval(coco_gt, iouType=iou_type)
+            self.coco_eval[iou_type].useCats = useCats
+
+        self.img_ids = []
+        self.eval_imgs = {k: [] for k in iou_types}
+        self.useCats = useCats
+
+    def update(self, predictions):
+        img_ids = list(np.unique(list(predictions.keys())))
+        self.img_ids.extend(img_ids)
+        # import pdb;pdb.set_trace()
+        for iou_type in self.iou_types:
+            results = self.prepare(predictions, iou_type)
+
+            # suppress pycocotools prints
+            with open(os.devnull, "w") as devnull:
+                with contextlib.redirect_stdout(devnull):
+                    coco_dt = COCO.loadRes(self.coco_gt, results) if results else COCO()
+
+            coco_eval = self.coco_eval[iou_type]
+
+            coco_eval.cocoDt = coco_dt
+            coco_eval.params.imgIds = list(img_ids)
+            coco_eval.params.useCats = self.useCats
+            img_ids, eval_imgs = evaluate(coco_eval)
+
+            self.eval_imgs[iou_type].append(eval_imgs)
+
+    def synchronize_between_processes(self):
+        for iou_type in self.iou_types:
+            self.eval_imgs[iou_type] = np.concatenate(self.eval_imgs[iou_type], 2)
+            create_common_coco_eval(self.coco_eval[iou_type], self.img_ids, self.eval_imgs[iou_type])
+
+    def accumulate(self):
+        for coco_eval in self.coco_eval.values():
+            coco_eval.accumulate()
+
+    def summarize(self):
+        for iou_type, coco_eval in self.coco_eval.items():
+            print("IoU metric: {}".format(iou_type))
+            coco_eval.summarize()
+
+    def prepare(self, predictions, iou_type):
+        if iou_type == "bbox":
+            return self.prepare_for_coco_detection(predictions)
+        elif iou_type == "segm":
+            return self.prepare_for_coco_segmentation(predictions)
+        elif iou_type == "keypoints":
+            return self.prepare_for_coco_keypoint(predictions)
+        else:
+            raise ValueError("Unknown iou type {}".format(iou_type))
+
+    def prepare_for_coco_detection(self, predictions):
+        coco_results = []
+        for original_id, prediction in predictions.items():
+            if len(prediction) == 0:
+                continue
+
+            boxes = prediction["boxes"]
+            boxes = convert_to_xywh(boxes).tolist()
+            scores = prediction["scores"].tolist()
+            labels = prediction["labels"].tolist()
+
+            coco_results.extend(
+                [
+                    {
+                        "image_id": original_id,
+                        "category_id": labels[k],
+                        "bbox": box,
+                        "score": scores[k],
+                    }
+                    for k, box in enumerate(boxes)
+                ]
+            )
+        return coco_results
+
+    def prepare_for_coco_segmentation(self, predictions):
+        coco_results = []
+        for original_id, prediction in predictions.items():
+            if len(prediction) == 0:
+                continue
+
+            scores = prediction["scores"]
+            labels = prediction["labels"]
+            masks = prediction["masks"]
+
+            masks = masks > 0.5
+
+            scores = prediction["scores"].tolist()
+            labels = prediction["labels"].tolist()
+
+            rles = [
+                mask_util.encode(np.array(mask[0, :, :, np.newaxis], dtype=np.uint8, order="F"))[0] 
+                for mask in masks
+            ]
+            for rle in rles:
+                rle["counts"] = rle["counts"].decode("utf-8")
+
+            coco_results.extend(
+                [
+                    {
+                        "image_id": original_id,
+                        "category_id": labels[k],
+                        "segmentation": rle,
+                        "score": scores[k],
+                    }
+                    for k, rle in enumerate(rles)
+                ]
+            )
+        return coco_results
+
+    def prepare_for_coco_keypoint(self, predictions):
+        coco_results = []
+        for original_id, prediction in predictions.items():
+            if len(prediction) == 0:
+                continue
+
+            boxes = prediction["boxes"]
+            boxes = convert_to_xywh(boxes).tolist()
+            scores = prediction["scores"].tolist()
+            labels = prediction["labels"].tolist()
+            keypoints = prediction["keypoints"]
+            keypoints = keypoints.flatten(start_dim=1).tolist()
+
+            coco_results.extend(
+                [
+                    {
+                        "image_id": original_id,
+                        "category_id": labels[k],
+                        "keypoints": keypoint,
+                        "score": scores[k],
+                    }
+                    for k, keypoint in enumerate(keypoints)
+                ]
+            )
+        return coco_results
+
+
+def convert_to_xywh(boxes):
+    xmin, ymin, xmax, ymax = boxes.unbind(1)
+    return torch.stack((xmin, ymin, xmax - xmin, ymax - ymin), dim=1)
+
+
+def merge(img_ids, eval_imgs):
+    all_img_ids = all_gather(img_ids)
+    all_eval_imgs = all_gather(eval_imgs)
+
+    merged_img_ids = []
+    for p in all_img_ids:
+        merged_img_ids.extend(p)
+
+    merged_eval_imgs = []
+    for p in all_eval_imgs:
+        merged_eval_imgs.append(p)
+
+    merged_img_ids = np.array(merged_img_ids)
+    merged_eval_imgs = np.concatenate(merged_eval_imgs, 2)
+
+    # keep only unique (and in sorted order) images
+    merged_img_ids, idx = np.unique(merged_img_ids, return_index=True)
+    merged_eval_imgs = merged_eval_imgs[..., idx]
+
+    return merged_img_ids, merged_eval_imgs
+
+
+def create_common_coco_eval(coco_eval, img_ids, eval_imgs):
+    img_ids, eval_imgs = merge(img_ids, eval_imgs)
+    img_ids = list(img_ids)
+    eval_imgs = list(eval_imgs.flatten())
+
+    coco_eval.evalImgs = eval_imgs
+    coco_eval.params.imgIds = img_ids
+    coco_eval._paramsEval = copy.deepcopy(coco_eval.params)
+
+
+#################################################################
+# From pycocotools, just removed the prints and fixed
+# a Python3 bug about unicode not defined
+#################################################################
+
+
+def evaluate(self):
+    """
+    Run per image evaluation on given images and store results (a list of dict) in self.evalImgs
+    :return: None
+    """
+    # tic = time.time()
+    # print('Running per image evaluation...')
+
+    # import pdb;pdb.set_trace()
+    p = self.params
+    # add backward compatibility if useSegm is specified in params
+    if p.useSegm is not None:
+        p.iouType = "segm" if p.useSegm == 1 else "bbox"
+        print("useSegm (deprecated) is not None. Running {} evaluation".format(p.iouType))
+    # print('Evaluate annotation type *{}*'.format(p.iouType))
+    p.imgIds = list(np.unique(p.imgIds))
+    if p.useCats:
+        p.catIds = list(np.unique(p.catIds))
+    p.maxDets = sorted(p.maxDets)
+    self.params = p
+
+    self._prepare()
+    # loop through images, area range, max detection number
+    catIds = p.catIds if p.useCats else [-1]
+
+    if p.iouType == "segm" or p.iouType == "bbox":
+        computeIoU = self.computeIoU
+    elif p.iouType == "keypoints":
+        computeIoU = self.computeOks
+    self.ious = {
+        (imgId, catId): computeIoU(imgId, catId) 
+        for imgId in p.imgIds 
+        for catId in catIds}
+
+    evaluateImg = self.evaluateImg
+    maxDet = p.maxDets[-1]
+    evalImgs = [
+        evaluateImg(imgId, catId, areaRng, maxDet) 
+        for catId in catIds 
+        for areaRng in p.areaRng 
+        for imgId in p.imgIds
+    ]
+    # this is NOT in the pycocotools code, but could be done outside
+    evalImgs = np.asarray(evalImgs).reshape(len(catIds), len(p.areaRng), len(p.imgIds))
+    self._paramsEval = copy.deepcopy(self.params)
+    # toc = time.time()
+    # print('DONE (t={:0.2f}s).'.format(toc-tic))
+    return p.imgIds, evalImgs
+
+
+#################################################################
+# end of straight copy from pycocotools, just removing the prints
+#################################################################

datasets/data_util.py

@@ -0,0 +1,170 @@
+import os
+import os.path as osp
+import shutil
+import time
+import datetime
+
+import torch
+
+from util.slconfig import SLConfig
+
+class Error(OSError):
+    pass
+
+def slcopytree(src, dst, symlinks=False, ignore=None, copy_function=shutil.copyfile,
+             ignore_dangling_symlinks=False):
+    """
+    modified from shutil.copytree without copystat.
+    
+    Recursively copy a directory tree.
+
+    The destination directory must not already exist.
+    If exception(s) occur, an Error is raised with a list of reasons.
+
+    If the optional symlinks flag is true, symbolic links in the
+    source tree result in symbolic links in the destination tree; if
+    it is false, the contents of the files pointed to by symbolic
+    links are copied. If the file pointed by the symlink doesn't
+    exist, an exception will be added in the list of errors raised in
+    an Error exception at the end of the copy process.
+
+    You can set the optional ignore_dangling_symlinks flag to true if you
+    want to silence this exception. Notice that this has no effect on
+    platforms that don't support os.symlink.
+
+    The optional ignore argument is a callable. If given, it
+    is called with the `src` parameter, which is the directory
+    being visited by copytree(), and `names` which is the list of
+    `src` contents, as returned by os.listdir():
+
+        callable(src, names) -> ignored_names
+
+    Since copytree() is called recursively, the callable will be
+    called once for each directory that is copied. It returns a
+    list of names relative to the `src` directory that should
+    not be copied.
+
+    The optional copy_function argument is a callable that will be used
+    to copy each file. It will be called with the source path and the
+    destination path as arguments. By default, copy2() is used, but any
+    function that supports the same signature (like copy()) can be used.
+
+    """
+    errors = []
+    if os.path.isdir(src):
+        names = os.listdir(src)
+        if ignore is not None:
+            ignored_names = ignore(src, names)
+        else:
+            ignored_names = set()
+
+        os.makedirs(dst)
+        for name in names:
+            if name in ignored_names:
+                continue
+            srcname = os.path.join(src, name)
+            dstname = os.path.join(dst, name)
+            try:
+                if os.path.islink(srcname):
+                    linkto = os.readlink(srcname)
+                    if symlinks:
+                        # We can't just leave it to `copy_function` because legacy
+                        # code with a custom `copy_function` may rely on copytree
+                        # doing the right thing.
+                        os.symlink(linkto, dstname)
+                    else:
+                        # ignore dangling symlink if the flag is on
+                        if not os.path.exists(linkto) and ignore_dangling_symlinks:
+                            continue
+                        # otherwise let the copy occurs. copy2 will raise an error
+                        if os.path.isdir(srcname):
+                            slcopytree(srcname, dstname, symlinks, ignore,
+                                    copy_function)
+                        else:
+                            copy_function(srcname, dstname)
+                elif os.path.isdir(srcname):
+                    slcopytree(srcname, dstname, symlinks, ignore, copy_function)
+                else:
+                    # Will raise a SpecialFileError for unsupported file types
+                    copy_function(srcname, dstname)
+            # catch the Error from the recursive copytree so that we can
+            # continue with other files
+            except Error as err:
+                errors.extend(err.args[0])
+            except OSError as why:
+                errors.append((srcname, dstname, str(why)))
+    else:
+        copy_function(src, dst)
+
+    if errors:
+        raise Error(errors)
+    return dst
+
+def check_and_copy(src_path, tgt_path):
+    if os.path.exists(tgt_path):
+        return None
+
+    return slcopytree(src_path, tgt_path)
+
+
+def remove(srcpath):
+    if os.path.isdir(srcpath):
+        return shutil.rmtree(srcpath)
+    else:
+        return os.remove(srcpath)  
+
+
+def preparing_dataset(pathdict, image_set, args):
+    start_time = time.time()
+    dataset_file = args.dataset_file
+    data_static_info = SLConfig.fromfile('util/static_data_path.py')
+    static_dict = data_static_info[dataset_file][image_set]
+
+    copyfilelist = []
+    for k,tgt_v in pathdict.items():
+        if os.path.exists(tgt_v):
+            if args.local_rank == 0:
+                print("path <{}> exist. remove it!".format(tgt_v))
+                remove(tgt_v)
+            # continue
+        
+        if args.local_rank == 0:
+            src_v = static_dict[k]
+            assert isinstance(src_v, str)
+            if src_v.endswith('.zip'):
+                # copy
+                cp_tgt_dir = os.path.dirname(tgt_v)
+                filename = os.path.basename(src_v)
+                cp_tgt_path = os.path.join(cp_tgt_dir, filename)
+                print('Copy from <{}> to <{}>.'.format(src_v, cp_tgt_path))
+                os.makedirs(cp_tgt_dir, exist_ok=True)
+                check_and_copy(src_v, cp_tgt_path)          
+
+                # unzip
+                import zipfile
+                print("Starting unzip <{}>".format(cp_tgt_path))
+                with zipfile.ZipFile(cp_tgt_path, 'r') as zip_ref:
+                    zip_ref.extractall(os.path.dirname(cp_tgt_path))      
+
+                copyfilelist.append(cp_tgt_path)
+                copyfilelist.append(tgt_v)
+            else:
+                print('Copy from <{}> to <{}>.'.format(src_v, tgt_v))
+                os.makedirs(os.path.dirname(tgt_v), exist_ok=True)
+                check_and_copy(src_v, tgt_v)
+                copyfilelist.append(tgt_v)
+    
+    if len(copyfilelist) == 0:
+        copyfilelist = None
+    args.copyfilelist = copyfilelist
+        
+    if args.distributed:
+        torch.distributed.barrier()
+    total_time = time.time() - start_time
+    if copyfilelist:
+        total_time_str = str(datetime.timedelta(seconds=int(total_time)))
+        print('Data copy time {}'.format(total_time_str))
+    return copyfilelist
+
+
+    

datasets/dataset.py

@@ -0,0 +1,44 @@
+from __future__ import print_function
+
+import torch
+import torchvision.datasets as datasets
+from torch.utils.data import Dataset
+from PIL import Image
+from .tsv_io import TSVFile
+import numpy as np
+import base64
+import io
+
+
+class TSVDataset(Dataset):
+    """ TSV dataset for ImageNet 1K training
+    """    
+    def __init__(self, tsv_file, transform=None, target_transform=None):
+        self.tsv = TSVFile(tsv_file)
+        self.transform = transform
+        self.target_transform = target_transform
+
+    def __getitem__(self, index):
+        """
+        Args:
+            index (int): Index
+        Returns:
+            tuple: (image, target) where target is class_index of the target class.
+        """
+        row = self.tsv.seek(index)
+        image_data = base64.b64decode(row[-1])
+        image = Image.open(io.BytesIO(image_data))
+        image = image.convert('RGB')
+        target = int(row[1])
+
+        if self.transform is not None:
+            img = self.transform(image)
+        else:
+            img = image
+        if self.target_transform is not None:
+            target = self.target_transform(target)
+
+        return img, target
+
+    def __len__(self):
+        return self.tsv.num_rows()

datasets/odvg.py

@@ -0,0 +1,288 @@
+from torchvision.datasets.vision import VisionDataset
+import os.path
+from typing import Callable, Optional
+import json
+from PIL import Image
+import torch
+import random
+import os, sys
+
+sys.path.append(os.path.dirname(sys.path[0]))
+
+import datasets.transforms as T
+
+
+class ODVGDataset(VisionDataset):
+    """
+    Args:
+        root (string): Root directory where images are downloaded to.
+        anno (string): Path to json annotation file.
+        label_map_anno (string):  Path to json label mapping file. Only for Object Detection
+        transform (callable, optional): A function/transform that  takes in an PIL image
+            and returns a transformed version. E.g, ``transforms.PILToTensor``
+        target_transform (callable, optional): A function/transform that takes in the
+            target and transforms it.
+        transforms (callable, optional): A function/transform that takes input sample and its target as entry
+            and returns a transformed version.
+    """
+
+    def __init__(
+        self,
+        root: str,
+        anno: str,
+        label_map_anno: str = None,
+        max_labels: int = 80,
+        transform: Optional[Callable] = None,
+        target_transform: Optional[Callable] = None,
+        transforms: Optional[Callable] = None,
+    ) -> None:
+        super().__init__(root, transforms, transform, target_transform)
+        self.root = root
+        self.dataset_mode = "OD" if label_map_anno else "VG"
+        self.max_labels = max_labels
+        if self.dataset_mode == "OD":
+            self.load_label_map(label_map_anno)
+        self._load_metas(anno)
+        self.get_dataset_info()
+
+    def load_label_map(self, label_map_anno):
+        with open(label_map_anno, "r") as file:
+            self.label_map = json.load(file)
+        self.label_index = set(self.label_map.keys())
+
+    def _load_metas(self, anno):
+        with open(anno, "r") as f:
+            self.metas = [json.loads(line) for line in f]
+
+    def get_dataset_info(self):
+        print(f"  == total images: {len(self)}")
+        if self.dataset_mode == "OD":
+            print(f"  == total labels: {len(self.label_map)}")
+
+    def __getitem__(self, index: int):
+        meta = self.metas[index]
+        rel_path = meta["filename"]
+        abs_path = os.path.join(self.root, rel_path)
+        if not os.path.exists(abs_path):
+            raise FileNotFoundError(f"{abs_path} not found.")
+        image = Image.open(abs_path).convert("RGB")
+        exemplars = torch.tensor(meta["exemplars"], dtype=torch.int64)
+        w, h = image.size
+        if self.dataset_mode == "OD":
+            anno = meta["detection"]
+            instances = [obj for obj in anno["instances"]]
+            boxes = [obj["bbox"] for obj in instances]
+            # generate vg_labels
+            # pos bbox labels
+            ori_classes = [str(obj["label"]) for obj in instances]
+            pos_labels = set(ori_classes)
+            # neg bbox labels
+            neg_labels = self.label_index.difference(pos_labels)
+
+            vg_labels = list(pos_labels)
+            num_to_add = min(len(neg_labels), self.max_labels - len(pos_labels))
+            if num_to_add > 0:
+                vg_labels.extend(random.sample(neg_labels, num_to_add))
+
+            # shuffle
+            for i in range(len(vg_labels) - 1, 0, -1):
+                j = random.randint(0, i)
+                vg_labels[i], vg_labels[j] = vg_labels[j], vg_labels[i]
+
+            caption_list = [self.label_map[lb] for lb in vg_labels]
+            caption_dict = {item: index for index, item in enumerate(caption_list)}
+
+            caption = " . ".join(caption_list) + " ."
+            classes = [
+                caption_dict[self.label_map[str(obj["label"])]] for obj in instances
+            ]
+            boxes = torch.as_tensor(boxes, dtype=torch.float32).reshape(-1, 4)
+            classes = torch.tensor(classes, dtype=torch.int64)
+        elif self.dataset_mode == "VG":
+            anno = meta["grounding"]
+            instances = [obj for obj in anno["regions"]]
+            boxes = [obj["bbox"] for obj in instances]
+            caption_list = [obj["phrase"] for obj in instances]
+            c = list(zip(boxes, caption_list))
+            random.shuffle(c)
+            boxes[:], caption_list[:] = zip(*c)
+            uni_caption_list = list(set(caption_list))
+            label_map = {}
+            for idx in range(len(uni_caption_list)):
+                label_map[uni_caption_list[idx]] = idx
+            classes = [label_map[cap] for cap in caption_list]
+            caption = " . ".join(uni_caption_list) + " ."
+            boxes = torch.as_tensor(boxes, dtype=torch.float32).reshape(-1, 4)
+            classes = torch.tensor(classes, dtype=torch.int64)
+            caption_list = uni_caption_list
+        target = {}
+        target["size"] = torch.as_tensor([int(h), int(w)])
+        target["cap_list"] = caption_list
+        target["caption"] = caption
+        target["boxes"] = boxes
+        target["labels"] = classes
+        target["exemplars"] = exemplars
+        target["labels_uncropped"] = torch.clone(classes)
+        # size, cap_list, caption, bboxes, labels
+
+        if self.transforms is not None:
+            image, target = self.transforms(image, target)
+        # Check that transforms does not change the identity of target['labels'].
+        if len(target["labels"]) > 0:
+            assert target["labels"][0] == target["labels_uncropped"][0]
+            print(
+                "Asserted that transforms does not change the identity of target['labels']."
+            )
+
+        return image, target
+
+    def __len__(self) -> int:
+        return len(self.metas)
+
+
+def make_coco_transforms(image_set, fix_size=False, strong_aug=False, args=None):
+    normalize = T.Compose(
+        [T.ToTensor(), T.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])]
+    )
+
+    # config the params for data aug
+    scales = [480, 512, 544, 576, 608, 640, 672, 704, 736, 768, 800]
+    max_size = 1333
+    scales2_resize = [400, 500, 600]
+    scales2_crop = [384, 600]
+
+    # update args from config files
+    scales = getattr(args, "data_aug_scales", scales)
+    max_size = getattr(args, "data_aug_max_size", max_size)
+    scales2_resize = getattr(args, "data_aug_scales2_resize", scales2_resize)
+    scales2_crop = getattr(args, "data_aug_scales2_crop", scales2_crop)
+
+    # resize them
+    data_aug_scale_overlap = getattr(args, "data_aug_scale_overlap", None)
+    if data_aug_scale_overlap is not None and data_aug_scale_overlap > 0:
+        data_aug_scale_overlap = float(data_aug_scale_overlap)
+        scales = [int(i * data_aug_scale_overlap) for i in scales]
+        max_size = int(max_size * data_aug_scale_overlap)
+        scales2_resize = [int(i * data_aug_scale_overlap) for i in scales2_resize]
+        scales2_crop = [int(i * data_aug_scale_overlap) for i in scales2_crop]
+
+    # datadict_for_print = {
+    #     'scales': scales,
+    #     'max_size': max_size,
+    #     'scales2_resize': scales2_resize,
+    #     'scales2_crop': scales2_crop
+    # }
+    # print("data_aug_params:", json.dumps(datadict_for_print, indent=2))
+
+    if image_set == "train":
+        if fix_size:
+            return T.Compose(
+                [
+                    T.RandomHorizontalFlip(),
+                    T.RandomResize([(max_size, max(scales))]),
+                    normalize,
+                ]
+            )
+
+        if strong_aug:
+            import datasets.sltransform as SLT
+
+            return T.Compose(
+                [
+                    T.RandomHorizontalFlip(),
+                    T.RandomSelect(
+                        T.RandomResize(scales, max_size=max_size),
+                        T.Compose(
+                            [
+                                T.RandomResize(scales2_resize),
+                                T.RandomSizeCrop(*scales2_crop),
+                                T.RandomResize(scales, max_size=max_size),
+                            ]
+                        ),
+                    ),
+                    SLT.RandomSelectMulti(
+                        [
+                            SLT.RandomCrop(),
+                            SLT.LightingNoise(),
+                            SLT.AdjustBrightness(2),
+                            SLT.AdjustContrast(2),
+                        ]
+                    ),
+                    normalize,
+                ]
+            )
+
+        return T.Compose(
+            [
+                T.RandomHorizontalFlip(),
+                T.RandomSelect(
+                    T.RandomResize(scales, max_size=max_size),
+                    T.Compose(
+                        [
+                            T.RandomResize(scales2_resize),
+                            T.RandomSizeCrop(*scales2_crop),
+                            T.RandomResize(scales, max_size=max_size),
+                        ]
+                    ),
+                ),
+                normalize,
+            ]
+        )
+
+    if image_set in ["val", "eval_debug", "train_reg", "test"]:
+        if os.environ.get("GFLOPS_DEBUG_SHILONG", False) == "INFO":
+            print("Under debug mode for flops calculation only!!!!!!!!!!!!!!!!")
+            return T.Compose(
+                [
+                    T.ResizeDebug((1280, 800)),
+                    normalize,
+                ]
+            )
+
+        return T.Compose(
+            [
+                T.RandomResize([max(scales)], max_size=max_size),
+                normalize,
+            ]
+        )
+
+    raise ValueError(f"unknown {image_set}")
+
+
+def build_odvg(image_set, args, datasetinfo):
+    img_folder = datasetinfo["root"]
+    ann_file = datasetinfo["anno"]
+    label_map = datasetinfo["label_map"] if "label_map" in datasetinfo else None
+    try:
+        strong_aug = args.strong_aug
+    except:
+        strong_aug = False
+    print(img_folder, ann_file, label_map)
+    dataset = ODVGDataset(
+        img_folder,
+        ann_file,
+        label_map,
+        max_labels=args.max_labels,
+        transforms=make_coco_transforms(
+            image_set, fix_size=args.fix_size, strong_aug=strong_aug, args=args
+        ),
+    )
+    return dataset
+
+
+if __name__ == "__main__":
+    dataset_vg = ODVGDataset(
+        "path/GRIT-20M/data/",
+        "path/GRIT-20M/anno/grit_odvg_10k.jsonl",
+    )
+    print(len(dataset_vg))
+    data = dataset_vg[random.randint(0, 100)]
+    print(data)
+    dataset_od = ODVGDataset(
+        "pathl/V3Det/",
+        "path/V3Det/annotations/v3det_2023_v1_all_odvg.jsonl",
+        "path/V3Det/annotations/v3det_label_map.json",
+    )
+    print(len(dataset_od))
+    data = dataset_od[random.randint(0, 100)]
+    print(data)

datasets/panoptic_eval.py

@@ -0,0 +1,44 @@
+# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved
+import json
+import os
+
+import util.misc as utils
+
+try:
+    from panopticapi.evaluation import pq_compute
+except ImportError:
+    pass
+
+
+class PanopticEvaluator(object):
+    def __init__(self, ann_file, ann_folder, output_dir="panoptic_eval"):
+        self.gt_json = ann_file
+        self.gt_folder = ann_folder
+        if utils.is_main_process():
+            if not os.path.exists(output_dir):
+                os.mkdir(output_dir)
+        self.output_dir = output_dir
+        self.predictions = []
+
+    def update(self, predictions):
+        for p in predictions:
+            with open(os.path.join(self.output_dir, p["file_name"]), "wb") as f:
+                f.write(p.pop("png_string"))
+
+        self.predictions += predictions
+
+    def synchronize_between_processes(self):
+        all_predictions = utils.all_gather(self.predictions)
+        merged_predictions = []
+        for p in all_predictions:
+            merged_predictions += p
+        self.predictions = merged_predictions
+
+    def summarize(self):
+        if utils.is_main_process():
+            json_data = {"annotations": self.predictions}
+            predictions_json = os.path.join(self.output_dir, "predictions.json")
+            with open(predictions_json, "w") as f:
+                f.write(json.dumps(json_data))
+            return pq_compute(self.gt_json, predictions_json, gt_folder=self.gt_folder, pred_folder=self.output_dir)
+        return None

datasets/random_crop.py

@@ -0,0 +1,135 @@
+import PIL #version 1.2.0
+import torch
+import os
+import torchvision.transforms.functional as F
+import numpy as np
+import random
+
+
+def intersect(boxes1, boxes2):
+    '''
+        Find intersection of every box combination between two sets of box
+        boxes1: bounding boxes 1, a tensor of dimensions (n1, 4)
+        boxes2: bounding boxes 2, a tensor of dimensions (n2, 4)
+        
+        Out: Intersection each of boxes1 with respect to each of boxes2, 
+             a tensor of dimensions (n1, n2)
+    '''
+    n1 = boxes1.size(0)
+    n2 = boxes2.size(0)
+    max_xy =  torch.min(boxes1[:, 2:].unsqueeze(1).expand(n1, n2, 2),
+                        boxes2[:, 2:].unsqueeze(0).expand(n1, n2, 2))
+    
+    min_xy = torch.max(boxes1[:, :2].unsqueeze(1).expand(n1, n2, 2),
+                       boxes2[:, :2].unsqueeze(0).expand(n1, n2, 2))
+    inter = torch.clamp(max_xy - min_xy , min=0)  # (n1, n2, 2)
+    return inter[:, :, 0] * inter[:, :, 1]  #(n1, n2)
+def find_IoU(boxes1, boxes2):
+    '''
+        Find IoU between every boxes set of boxes 
+        boxes1: a tensor of dimensions (n1, 4) (left, top, right , bottom)
+        boxes2: a tensor of dimensions (n2, 4)
+        
+        Out: IoU each of boxes1 with respect to each of boxes2, a tensor of 
+             dimensions (n1, n2)
+        
+        Formula: 
+        (box1 ∩ box2) / (box1 u box2) = (box1 ∩ box2) / (area(box1) + area(box2) - (box1 ∩ box2 ))
+    '''
+    inter = intersect(boxes1, boxes2)
+    area_boxes1 = (boxes1[:, 2] - boxes1[:, 0]) * (boxes1[:, 3] - boxes1[:, 1])
+    area_boxes2 = (boxes2[:, 2] - boxes2[:, 0]) * (boxes2[:, 3] - boxes2[:, 1])
+    
+    area_boxes1 = area_boxes1.unsqueeze(1).expand_as(inter) #(n1, n2)
+    area_boxes2 = area_boxes2.unsqueeze(0).expand_as(inter)  #(n1, n2)
+    union = (area_boxes1 + area_boxes2 - inter)
+    return inter / union
+
+
+def random_crop(image, boxes, labels, difficulties=None):
+    '''
+        image: A PIL image
+        boxes: Bounding boxes, a tensor of dimensions (#objects, 4)
+        labels: labels of object, a tensor of dimensions (#objects)
+        difficulties: difficulties of detect object, a tensor of dimensions (#objects)
+        
+        Out: cropped image , new boxes, new labels, new difficulties
+    '''
+    if type(image) == PIL.Image.Image:
+        image = F.to_tensor(image)
+    original_h = image.size(1)
+    original_w = image.size(2)
+    
+    while True:
+        mode = random.choice([0.1, 0.3, 0.5, 0.9, None])
+        
+        if mode is None:
+            return F.to_pil_image(image), boxes, labels, difficulties
+        
+        new_image = image
+        new_boxes = boxes
+        new_difficulties = difficulties
+        new_labels = labels
+        for _ in range(50):
+            # Crop dimensions: [0.3, 1] of original dimensions
+            new_h = random.uniform(0.3*original_h, original_h)
+            new_w = random.uniform(0.3*original_w, original_w)
+            
+            # Aspect ratio constraint b/t .5 & 2
+            if new_h/new_w < 0.5 or new_h/new_w > 2:
+                continue
+            
+            #Crop coordinate
+            left = random.uniform(0, original_w - new_w)
+            right = left + new_w
+            top = random.uniform(0, original_h - new_h)
+            bottom = top + new_h
+            crop = torch.FloatTensor([int(left), int(top), int(right), int(bottom)])
+            
+            # Calculate IoU  between the crop and the bounding boxes
+            overlap = find_IoU(crop.unsqueeze(0), boxes) #(1, #objects)
+            overlap = overlap.squeeze(0)
+
+            # If not a single bounding box has a IoU of greater than the minimum, try again
+            if overlap.shape[0] == 0:
+                continue
+            if overlap.max().item() < mode:
+                continue
+            
+            #Crop
+            new_image = image[:, int(top):int(bottom), int(left):int(right)] #(3, new_h, new_w)
+            
+            #Center of bounding boxes
+            center_bb = (boxes[:, :2] + boxes[:, 2:])/2.0
+            
+            #Find bounding box has been had center in crop
+            center_in_crop = (center_bb[:, 0] >left) * (center_bb[:, 0] < right
+                             ) *(center_bb[:, 1] > top) * (center_bb[:, 1] < bottom)    #( #objects)
+            
+            if not center_in_crop.any():
+                continue
+            
+            #take matching bounding box
+            new_boxes = boxes[center_in_crop, :]
+            
+            #take matching labels
+            new_labels = labels[center_in_crop]
+            
+            #take matching difficulities
+            if difficulties is not None:
+                new_difficulties = difficulties[center_in_crop]
+            else:
+                new_difficulties = None
+            
+            #Use the box left and top corner or the crop's
+            new_boxes[:, :2] = torch.max(new_boxes[:, :2], crop[:2])
+            
+            #adjust to crop
+            new_boxes[:, :2] -= crop[:2]
+            
+            new_boxes[:, 2:] = torch.min(new_boxes[:, 2:],crop[2:])
+            
+            #adjust to crop
+            new_boxes[:, 2:] -= crop[:2]
+            
+            return F.to_pil_image(new_image), new_boxes, new_labels, new_difficulties

datasets/sltransform.py

@@ -0,0 +1,247 @@
+# modified from https://github.com/anhtuan85/Data-Augmentation-for-Object-Detection/blob/master/augmentation.ipynb
+
+import PIL #version 1.2.0
+from PIL import Image #version 6.1.0
+import torch
+import os
+import torchvision.transforms.functional as F
+import numpy as np
+import random
+
+from .random_crop import random_crop
+from util.box_ops import box_cxcywh_to_xyxy, box_xyxy_to_cxcywh
+
+class AdjustContrast:
+    def __init__(self, contrast_factor):
+        self.contrast_factor = contrast_factor
+
+    def __call__(self, img, target):
+        """
+        img (PIL Image or Tensor): Image to be adjusted.
+        """
+        _contrast_factor = ((random.random() + 1.0) / 2.0) * self.contrast_factor
+        img = F.adjust_contrast(img, _contrast_factor)
+        return img, target
+
+class AdjustBrightness:
+    def __init__(self, brightness_factor):
+        self.brightness_factor = brightness_factor
+
+    def __call__(self, img, target):
+        """
+        img (PIL Image or Tensor): Image to be adjusted.
+        """
+        _brightness_factor = ((random.random() + 1.0) / 2.0) * self.brightness_factor
+        img = F.adjust_brightness(img, _brightness_factor)
+        return img, target
+
+def lighting_noise(image):
+    '''
+        color channel swap in image
+        image: A PIL image
+    '''
+    new_image = image
+    perms = ((0, 1, 2), (0, 2, 1), (1, 0, 2), 
+             (1, 2, 0), (2, 0, 1), (2, 1, 0))
+    swap = perms[random.randint(0, len(perms)- 1)]
+    new_image = F.to_tensor(new_image)
+    new_image = new_image[swap, :, :]
+    new_image = F.to_pil_image(new_image)
+    return new_image
+
+class LightingNoise:
+    def __init__(self) -> None:
+        pass
+
+    def __call__(self, img, target):
+        return lighting_noise(img), target
+
+
+def rotate(image, boxes, angle):
+    '''
+        Rotate image and bounding box
+        image: A Pil image (w, h)
+        boxes: A tensors of dimensions (#objects, 4)
+        
+        Out: rotated image (w, h), rotated boxes
+    '''
+    new_image = image.copy()
+    new_boxes = boxes.clone()
+    
+    #Rotate image, expand = True
+    w = image.width
+    h = image.height
+    cx = w/2
+    cy = h/2
+    new_image = new_image.rotate(angle, expand=True)
+    angle = np.radians(angle)
+    alpha = np.cos(angle)
+    beta = np.sin(angle)
+    #Get affine matrix
+    AffineMatrix = torch.tensor([[alpha, beta, (1-alpha)*cx - beta*cy],
+                                 [-beta, alpha, beta*cx + (1-alpha)*cy]])
+    
+    #Rotation boxes
+    box_width = (boxes[:,2] - boxes[:,0]).reshape(-1,1)
+    box_height = (boxes[:,3] - boxes[:,1]).reshape(-1,1)
+    
+    #Get corners for boxes
+    x1 = boxes[:,0].reshape(-1,1)
+    y1 = boxes[:,1].reshape(-1,1)
+    
+    x2 = x1 + box_width
+    y2 = y1 
+    
+    x3 = x1
+    y3 = y1 + box_height
+    
+    x4 = boxes[:,2].reshape(-1,1)
+    y4 = boxes[:,3].reshape(-1,1)
+    
+    corners = torch.stack((x1,y1,x2,y2,x3,y3,x4,y4), dim= 1)
+    # corners.reshape(-1, 8)    #Tensors of dimensions (#objects, 8)
+    corners = corners.reshape(-1,2) #Tensors of dimension (4* #objects, 2)
+    corners = torch.cat((corners, torch.ones(corners.shape[0], 1)), dim= 1) #(Tensors of dimension (4* #objects, 3))
+    
+    cos = np.abs(AffineMatrix[0, 0])
+    sin = np.abs(AffineMatrix[0, 1])
+    
+    nW = int((h * sin) + (w * cos))
+    nH = int((h * cos) + (w * sin))
+    AffineMatrix[0, 2] += (nW / 2) - cx
+    AffineMatrix[1, 2] += (nH / 2) - cy
+    
+
+    #Apply affine transform
+    rotate_corners = torch.mm(AffineMatrix, corners.t().to(torch.float64)).t()
+    rotate_corners = rotate_corners.reshape(-1,8)
+    
+    x_corners = rotate_corners[:,[0,2,4,6]]
+    y_corners = rotate_corners[:,[1,3,5,7]]
+    
+    #Get (x_min, y_min, x_max, y_max)
+    x_min, _ = torch.min(x_corners, dim= 1)
+    x_min = x_min.reshape(-1, 1)
+    y_min, _ = torch.min(y_corners, dim= 1)
+    y_min = y_min.reshape(-1, 1)
+    x_max, _ = torch.max(x_corners, dim= 1)
+    x_max = x_max.reshape(-1, 1)
+    y_max, _ = torch.max(y_corners, dim= 1)
+    y_max = y_max.reshape(-1, 1)
+    
+    new_boxes = torch.cat((x_min, y_min, x_max, y_max), dim= 1)
+    
+    scale_x = new_image.width / w
+    scale_y = new_image.height / h
+    
+    #Resize new image to (w, h)
+
+    new_image = new_image.resize((w, h))
+    
+    #Resize boxes
+    new_boxes /= torch.Tensor([scale_x, scale_y, scale_x, scale_y])
+    new_boxes[:, 0] = torch.clamp(new_boxes[:, 0], 0, w)
+    new_boxes[:, 1] = torch.clamp(new_boxes[:, 1], 0, h)
+    new_boxes[:, 2] = torch.clamp(new_boxes[:, 2], 0, w)
+    new_boxes[:, 3] = torch.clamp(new_boxes[:, 3], 0, h)
+    return new_image, new_boxes
+
+# def convert_xywh_to_xyxy(boxes: torch.Tensor):
+#     _boxes = boxes.clone()
+#     box_xy = _boxes[:, :2]
+#     box_wh = _boxes[:, 2:]
+#     box_x1y1 = box_xy - box_wh/2 
+#     box_x2y2 = box_xy + box_wh/2
+#     box_xyxy = torch.cat((box_x1y1, box_x2y2), dim=-1)
+#     return box_xyxy
+
+class Rotate:
+    def __init__(self, angle=10) -> None:
+        self.angle = angle
+
+    def __call__(self, img, target):
+        w,h = img.size
+        whwh = torch.Tensor([w, h, w, h])
+        boxes_xyxy = box_cxcywh_to_xyxy(target['boxes']) * whwh
+        img, boxes_new = rotate(img, boxes_xyxy, self.angle)
+        target['boxes'] = box_xyxy_to_cxcywh(boxes_new).to(boxes_xyxy.dtype) / (whwh + 1e-3)
+        return img, target
+
+
+class RandomCrop:
+    def __init__(self) -> None:
+        pass
+
+    def __call__(self, img, target):
+        w,h = img.size
+        try:
+            boxes_xyxy = target['boxes']
+            labels = target['labels']
+            img, new_boxes, new_labels, _ = random_crop(img, boxes_xyxy, labels)
+            target['boxes'] = new_boxes
+            target['labels'] = new_labels
+        except Exception as e:
+            pass
+        return img, target
+
+
+class RandomCropDebug:
+    def __init__(self) -> None:
+        pass
+
+    def __call__(self, img, target):
+        boxes_xyxy = target['boxes'].clone()
+        labels = target['labels'].clone()
+        img, new_boxes, new_labels, _ = random_crop(img, boxes_xyxy, labels)
+        target['boxes'] = new_boxes
+        target['labels'] = new_labels
+
+
+        return img, target
+        
+class RandomSelectMulti(object):
+    """
+    Randomly selects between transforms1 and transforms2,
+    """
+    def __init__(self, transformslist, p=-1):
+        self.transformslist = transformslist
+        self.p = p
+        assert p == -1
+
+    def __call__(self, img, target):
+        if self.p == -1:
+            return random.choice(self.transformslist)(img, target)
+
+
+class Albumentations:
+    def __init__(self):
+        import albumentations as A
+        self.transform = A.Compose([
+            A.Blur(p=0.01),
+            A.MedianBlur(p=0.01),
+            A.ToGray(p=0.01),
+            A.CLAHE(p=0.01),
+            A.RandomBrightnessContrast(p=0.005),
+            A.RandomGamma(p=0.005),
+            A.ImageCompression(quality_lower=75, p=0.005)],
+            bbox_params=A.BboxParams(format='pascal_voc', label_fields=['class_labels']))
+
+    def __call__(self, img, target, p=1.0):
+        """
+        Input:
+            target['boxes']: xyxy, unnormalized data.
+        
+        """
+        boxes_raw = target['boxes']
+        labels_raw = target['labels']
+        img_np = np.array(img)
+        if self.transform and random.random() < p:
+            new_res = self.transform(image=img_np, bboxes=boxes_raw, class_labels=labels_raw)  # transformed
+            boxes_new = torch.Tensor(new_res['bboxes']).to(boxes_raw.dtype).reshape_as(boxes_raw)
+            img_np = new_res['image']
+            labels_new = torch.Tensor(new_res['class_labels']).to(labels_raw.dtype)
+        img_new = Image.fromarray(img_np)
+        target['boxes'] = boxes_new
+        target['labels'] = labels_new
+        
+        return img_new, target

datasets/transforms.py

@@ -0,0 +1,338 @@
+# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved
+"""
+Transforms and data augmentation for both image + bbox.
+"""
+import random
+
+import PIL
+import torch
+import torchvision.transforms as T
+import torchvision.transforms.functional as F
+
+from util.box_ops import box_xyxy_to_cxcywh
+from util.misc import interpolate
+
+
+def crop(image, target, region):
+    cropped_image = F.crop(image, *region)
+
+    target = target.copy()
+    i, j, h, w = region
+
+    # should we do something wrt the original size?
+    target["size"] = torch.tensor([h, w])
+
+    fields = ["labels", "area"]
+
+    # Crop exemplars.
+    exemplars = target["exemplars"]
+    max_size = torch.as_tensor([w, h], dtype=torch.float32)
+    # Shift exemplars to cropped region.
+    cropped_exemplars = exemplars - torch.as_tensor([j, i, j, i])
+    # Correct exemplar regions that go past new image boundary (too far right).
+    cropped_exemplars = torch.min(cropped_exemplars.reshape(-1, 2, 2), max_size)
+    # Correct exemplar regions that go past new image boundary (too far left).
+    cropped_exemplars = cropped_exemplars.clamp(min=0)
+    # Get new exemplar areas.
+    area_exemplars = (cropped_exemplars[:, 1, :] - cropped_exemplars[:, 0, :]).prod(
+        dim=1
+    )
+    # Update [target] with cropped exemplars.
+    target["exemplars"] = cropped_exemplars.reshape(-1, 4)
+    if "boxes" in target:
+        boxes = target["boxes"]
+        max_size = torch.as_tensor([w, h], dtype=torch.float32)
+        cropped_boxes = boxes - torch.as_tensor([j, i, j, i])
+        cropped_boxes = torch.min(cropped_boxes.reshape(-1, 2, 2), max_size)
+        cropped_boxes = cropped_boxes.clamp(min=0)
+        area = (cropped_boxes[:, 1, :] - cropped_boxes[:, 0, :]).prod(dim=1)
+        target["boxes"] = cropped_boxes.reshape(-1, 4)
+        target["area"] = area
+        fields.append("boxes")
+
+    if "masks" in target:
+        # FIXME should we update the area here if there are no boxes?
+        target["masks"] = target["masks"][:, i : i + h, j : j + w]
+        fields.append("masks")
+
+    # Remove exemplars that have zero area (due to cropping).
+    keep = area_exemplars > 0
+    target["exemplars"] = target["exemplars"][keep, :]
+
+    # remove elements for which the boxes or masks that have zero area
+    if "boxes" in target or "masks" in target:
+        # favor boxes selection when defining which elements to keep
+        # this is compatible with previous implementation
+        if "boxes" in target:
+            cropped_boxes = target["boxes"].reshape(-1, 2, 2)
+            keep = torch.all(cropped_boxes[:, 1, :] > cropped_boxes[:, 0, :], dim=1)
+        else:
+            keep = target["masks"].flatten(1).any(1)
+
+        for field in fields:
+            target[field] = target[field][keep]
+
+    return cropped_image, target
+
+
+def hflip(image, target):
+    flipped_image = F.hflip(image)
+
+    w, h = image.size
+
+    target = target.copy()
+
+    exemplars = target["exemplars"]
+    # Flip image across x-axis.
+    exemplars = exemplars[:, [2, 1, 0, 3]] * torch.as_tensor([-1, 1, -1, 1])
+    # Shift flipped image to (0, 0).
+    exemplars = exemplars + torch.as_tensor([w, 0, w, 0])
+    # Update [target] with horizontally flipped exemplars.
+    target["exemplars"] = exemplars
+
+    if "boxes" in target:
+        boxes = target["boxes"]
+        boxes = boxes[:, [2, 1, 0, 3]] * torch.as_tensor(
+            [-1, 1, -1, 1]
+        ) + torch.as_tensor([w, 0, w, 0])
+        target["boxes"] = boxes
+
+    if "masks" in target:
+        target["masks"] = target["masks"].flip(-1)
+
+    return flipped_image, target
+
+
+def resize(image, target, size, max_size=None):
+    # size can be min_size (scalar) or (w, h) tuple
+
+    def get_size_with_aspect_ratio(image_size, size, max_size=None):
+        w, h = image_size
+        if max_size is not None:
+            min_original_size = float(min((w, h)))
+            max_original_size = float(max((w, h)))
+            if max_original_size / min_original_size * size > max_size:
+                size = int(round(max_size * min_original_size / max_original_size))
+
+        if (w <= h and w == size) or (h <= w and h == size):
+            return (h, w)
+
+        if w < h:
+            ow = size
+            oh = int(size * h / w)
+        else:
+            oh = size
+            ow = int(size * w / h)
+
+        return (oh, ow)
+
+    def get_size(image_size, size, max_size=None):
+        if isinstance(size, (list, tuple)):
+            return size[::-1]
+        else:
+            return get_size_with_aspect_ratio(image_size, size, max_size)
+
+    try:
+        size = get_size(image.size, size, max_size)
+    except:
+        size = get_size((image.shape[-1], image.shape[-2]), size, max_size)
+    rescaled_image = F.resize(image, size)
+
+    if target is None:
+        return rescaled_image, None
+
+    ratios = tuple(
+        float(s) / float(s_orig) for s, s_orig in zip(rescaled_image.size, image.size)
+    )
+    ratio_width, ratio_height = ratios
+
+    target = target.copy()
+
+    # Rescale exemplars.
+    exemplars = target["exemplars"]
+    if exemplars.shape[-1] == 4:
+        scaled_exemplars = exemplars * torch.as_tensor(
+            [ratio_width, ratio_height, ratio_width, ratio_height]
+        )
+    else:
+        scaled_exemplars = exemplars
+
+    target["exemplars"] = scaled_exemplars
+
+    if "boxes" in target:
+        boxes = target["boxes"]
+        scaled_boxes = boxes * torch.as_tensor(
+            [ratio_width, ratio_height, ratio_width, ratio_height]
+        )
+        target["boxes"] = scaled_boxes
+
+    if "area" in target:
+        area = target["area"]
+        scaled_area = area * (ratio_width * ratio_height)
+        target["area"] = scaled_area
+
+    h, w = size
+    target["size"] = torch.tensor([h, w])
+
+    if "masks" in target:
+        target["masks"] = (
+            interpolate(target["masks"][:, None].float(), size, mode="nearest")[:, 0]
+            > 0.5
+        )
+
+    return rescaled_image, target
+
+
+def pad(image, target, padding):
+    # assumes that we only pad on the bottom right corners
+    padded_image = F.pad(image, (0, 0, padding[0], padding[1]))
+    if target is None:
+        return padded_image, None
+    target = target.copy()
+    # should we do something wrt the original size?
+    target["size"] = torch.tensor(padded_image.size[::-1])
+    if "masks" in target:
+        target["masks"] = torch.nn.functional.pad(
+            target["masks"], (0, padding[0], 0, padding[1])
+        )
+    return padded_image, target
+
+
+class ResizeDebug(object):
+    def __init__(self, size):
+        self.size = size
+
+    def __call__(self, img, target):
+        return resize(img, target, self.size)
+
+
+class RandomCrop(object):
+    def __init__(self, size):
+        self.size = size
+
+    def __call__(self, img, target):
+        region = T.RandomCrop.get_params(img, self.size)
+        return crop(img, target, region)
+
+
+class RandomSizeCrop(object):
+    def __init__(self, min_size: int, max_size: int):
+        self.min_size = min_size
+        self.max_size = max_size
+
+    def __call__(self, img: PIL.Image.Image, target: dict):
+        w = random.randint(self.min_size, min(img.width, self.max_size))
+        h = random.randint(self.min_size, min(img.height, self.max_size))
+        region = T.RandomCrop.get_params(img, [h, w])
+        return crop(img, target, region)
+
+
+class CenterCrop(object):
+    def __init__(self, size):
+        self.size = size
+
+    def __call__(self, img, target):
+        image_width, image_height = img.size
+        crop_height, crop_width = self.size
+        crop_top = int(round((image_height - crop_height) / 2.0))
+        crop_left = int(round((image_width - crop_width) / 2.0))
+        return crop(img, target, (crop_top, crop_left, crop_height, crop_width))
+
+
+class RandomHorizontalFlip(object):
+    def __init__(self, p=0.5):
+        self.p = p
+
+    def __call__(self, img, target):
+        if random.random() < self.p:
+            return hflip(img, target)
+        return img, target
+
+
+class RandomResize(object):
+    def __init__(self, sizes, max_size=None):
+        assert isinstance(sizes, (list, tuple))
+        self.sizes = sizes
+        self.max_size = max_size
+
+    def __call__(self, img, target=None):
+        size = random.choice(self.sizes)
+        return resize(img, target, size, self.max_size)
+
+
+class RandomPad(object):
+    def __init__(self, max_pad):
+        self.max_pad = max_pad
+
+    def __call__(self, img, target):
+        pad_x = random.randint(0, self.max_pad)
+        pad_y = random.randint(0, self.max_pad)
+        return pad(img, target, (pad_x, pad_y))
+
+
+class RandomSelect(object):
+    """
+    Randomly selects between transforms1 and transforms2,
+    with probability p for transforms1 and (1 - p) for transforms2
+    """
+
+    def __init__(self, transforms1, transforms2, p=0.5):
+        self.transforms1 = transforms1
+        self.transforms2 = transforms2
+        self.p = p
+
+    def __call__(self, img, target):
+        if random.random() < self.p:
+            return self.transforms1(img, target)
+        return self.transforms2(img, target)
+
+
+class ToTensor(object):
+    def __call__(self, img, target):
+        return F.to_tensor(img), target
+
+
+class RandomErasing(object):
+    def __init__(self, *args, **kwargs):
+        self.eraser = T.RandomErasing(*args, **kwargs)
+
+    def __call__(self, img, target):
+        return self.eraser(img), target
+
+
+class Normalize(object):
+    def __init__(self, mean, std):
+        self.mean = mean
+        self.std = std
+
+    def __call__(self, image, target=None):
+        image = F.normalize(image, mean=self.mean, std=self.std)
+        if target is None:
+            return image, None
+        target = target.copy()
+        h, w = image.shape[-2:]
+        # No normalization of exemplars needed, since they are used directly for cropping.
+        if "boxes" in target:
+            boxes = target["boxes"]
+            boxes = box_xyxy_to_cxcywh(boxes)
+            boxes = boxes / torch.tensor([w, h, w, h], dtype=torch.float32)
+            target["boxes"] = boxes
+        return image, target
+
+
+class Compose(object):
+    def __init__(self, transforms):
+        self.transforms = transforms
+
+    def __call__(self, image, target):
+        for t in self.transforms:
+            image, target = t(image, target)
+        return image, target
+
+    def __repr__(self):
+        format_string = self.__class__.__name__ + "("
+        for t in self.transforms:
+            format_string += "\n"
+            format_string += "    {0}".format(t)
+        format_string += "\n)"
+        return format_string

debug/config_args_all.json

@@ -0,0 +1,308 @@
+{
+  "config_file": "config/cfg_fsc147_vit_b.py",
+  "options": {
+    "text_encoder_type": "checkpoints/bert-base-uncased"
+  },
+  "datasets": "config/datasets_fsc147.json",
+  "remove_difficult": false,
+  "fix_size": false,
+  "output_dir": "./debug",
+  "note": "",
+  "device": "cuda",
+  "seed": 42,
+  "resume": "",
+  "pretrain_model_path": "checkpoints/groundingdino_swinb_cogcoor.pth",
+  "finetune_ignore": null,
+  "start_epoch": 0,
+  "eval": false,
+  "num_workers": 8,
+  "test": false,
+  "debug": false,
+  "find_unused_params": false,
+  "save_results": false,
+  "save_log": false,
+  "world_size": 1,
+  "dist_url": "env://",
+  "rank": 0,
+  "local_rank": 0,
+  "amp": false,
+  "distributed": false,
+  "data_aug_scales": [
+    480,
+    512,
+    544,
+    576,
+    608,
+    640,
+    672,
+    704,
+    736,
+    768,
+    800
+  ],
+  "data_aug_max_size": 1333,
+  "data_aug_scales2_resize": [
+    400,
+    500,
+    600
+  ],
+  "data_aug_scales2_crop": [
+    384,
+    600
+  ],
+  "data_aug_scale_overlap": null,
+  "batch_size": 4,
+  "modelname": "groundingdino",
+  "backbone": "swin_B_384_22k",
+  "position_embedding": "sine",
+  "pe_temperatureH": 20,
+  "pe_temperatureW": 20,
+  "return_interm_indices": [
+    1,
+    2,
+    3
+  ],
+  "enc_layers": 6,
+  "dec_layers": 6,
+  "pre_norm": false,
+  "dim_feedforward": 2048,
+  "hidden_dim": 256,
+  "dropout": 0.0,
+  "nheads": 8,
+  "num_queries": 900,
+  "query_dim": 4,
+  "num_patterns": 0,
+  "num_feature_levels": 4,
+  "enc_n_points": 4,
+  "dec_n_points": 4,
+  "two_stage_type": "standard",
+  "two_stage_bbox_embed_share": false,
+  "two_stage_class_embed_share": false,
+  "transformer_activation": "relu",
+  "dec_pred_bbox_embed_share": true,
+  "dn_box_noise_scale": 1.0,
+  "dn_label_noise_ratio": 0.5,
+  "dn_label_coef": 1.0,
+  "dn_bbox_coef": 1.0,
+  "embed_init_tgt": true,
+  "dn_labelbook_size": 91,
+  "max_text_len": 256,
+  "text_encoder_type": "checkpoints/bert-base-uncased",
+  "use_text_enhancer": true,
+  "use_fusion_layer": true,
+  "use_checkpoint": true,
+  "use_transformer_ckpt": true,
+  "use_text_cross_attention": true,
+  "text_dropout": 0.0,
+  "fusion_dropout": 0.0,
+  "fusion_droppath": 0.1,
+  "sub_sentence_present": true,
+  "max_labels": 90,
+  "lr": 0.0001,
+  "backbone_freeze_keywords": null,
+  "freeze_keywords": [
+    "backbone.0",
+    "bert"
+  ],
+  "lr_backbone": 1e-05,
+  "lr_backbone_names": [
+    "backbone.0",
+    "bert"
+  ],
+  "lr_linear_proj_mult": 1e-05,
+  "lr_linear_proj_names": [
+    "ref_point_head",
+    "sampling_offsets"
+  ],
+  "weight_decay": 0.0001,
+  "param_dict_type": "ddetr_in_mmdet",
+  "ddetr_lr_param": false,
+  "epochs": 30,
+  "lr_drop": 10,
+  "save_checkpoint_interval": 10,
+  "clip_max_norm": 0.1,
+  "onecyclelr": false,
+  "multi_step_lr": false,
+  "lr_drop_list": [
+    10,
+    20
+  ],
+  "frozen_weights": null,
+  "dilation": false,
+  "pdetr3_bbox_embed_diff_each_layer": false,
+  "pdetr3_refHW": -1,
+  "random_refpoints_xy": false,
+  "fix_refpoints_hw": -1,
+  "dabdetr_yolo_like_anchor_update": false,
+  "dabdetr_deformable_encoder": false,
+  "dabdetr_deformable_decoder": false,
+  "use_deformable_box_attn": false,
+  "box_attn_type": "roi_align",
+  "dec_layer_number": null,
+  "decoder_layer_noise": false,
+  "dln_xy_noise": 0.2,
+  "dln_hw_noise": 0.2,
+  "add_channel_attention": false,
+  "add_pos_value": false,
+  "two_stage_pat_embed": 0,
+  "two_stage_add_query_num": 0,
+  "two_stage_learn_wh": false,
+  "two_stage_default_hw": 0.05,
+  "two_stage_keep_all_tokens": false,
+  "num_select": 900,
+  "batch_norm_type": "FrozenBatchNorm2d",
+  "masks": false,
+  "aux_loss": true,
+  "set_cost_class": 5.0,
+  "set_cost_bbox": 1.0,
+  "set_cost_giou": 0.0,
+  "cls_loss_coef": 5.0,
+  "bbox_loss_coef": 1.0,
+  "giou_loss_coef": 0.0,
+  "enc_loss_coef": 1.0,
+  "interm_loss_coef": 1.0,
+  "no_interm_box_loss": false,
+  "mask_loss_coef": 1.0,
+  "dice_loss_coef": 1.0,
+  "focal_alpha": 0.25,
+  "focal_gamma": 2.0,
+  "decoder_sa_type": "sa",
+  "matcher_type": "HungarianMatcher",
+  "decoder_module_seq": [
+    "sa",
+    "ca",
+    "ffn"
+  ],
+  "nms_iou_threshold": -1,
+  "dec_pred_class_embed_share": true,
+  "match_unstable_error": true,
+  "use_detached_boxes_dec_out": false,
+  "dn_scalar": 100,
+  "box_threshold": 0.23,
+  "text_threshold": 0,
+  "use_coco_eval": false,
+  "label_list": [
+    "alcohol bottle",
+    "baguette roll",
+    "ball",
+    "banana",
+    "bead",
+    "bee",
+    "birthday candle",
+    "biscuit",
+    "boat",
+    "bottle",
+    "bowl",
+    "box",
+    "bread roll",
+    "brick",
+    "buffalo",
+    "bun",
+    "calamari ring",
+    "can",
+    "candle",
+    "cap",
+    "car",
+    "cartridge",
+    "cassette",
+    "cement bag",
+    "cereal",
+    "chewing gum piece",
+    "chopstick",
+    "clam",
+    "coffee bean",
+    "coin",
+    "cotton ball",
+    "cow",
+    "crane",
+    "crayon",
+    "croissant",
+    "crow",
+    "cup",
+    "cupcake",
+    "cupcake holder",
+    "fish",
+    "gemstone",
+    "go game piece",
+    "goat",
+    "goldfish snack",
+    "goose",
+    "ice cream",
+    "ice cream cone",
+    "instant noodle",
+    "jade stone",
+    "jeans",
+    "kidney bean",
+    "kitchen towel",
+    "lighter",
+    "lipstick",
+    "m&m piece",
+    "macaron",
+    "match",
+    "meat skewer",
+    "mini blind",
+    "mosaic tile",
+    "naan bread",
+    "nail",
+    "nut",
+    "onion ring",
+    "orange",
+    "pearl",
+    "pen",
+    "pencil",
+    "penguin",
+    "pepper",
+    "person",
+    "pigeon",
+    "plate",
+    "polka dot tile",
+    "potato",
+    "rice bag",
+    "roof tile",
+    "screw",
+    "shoe",
+    "spoon",
+    "spring roll",
+    "stair",
+    "stapler pin",
+    "straw",
+    "supermarket shelf",
+    "swan",
+    "tomato",
+    "watermelon",
+    "window",
+    "zebra"
+  ],
+  "val_label_list": [
+    "ant",
+    "bird",
+    "book",
+    "bottle cap",
+    "bullet",
+    "camel",
+    "chair",
+    "chicken wing",
+    "donut",
+    "donut holder",
+    "flamingo",
+    "flower",
+    "flower pot",
+    "grape",
+    "horse",
+    "kiwi",
+    "milk carton",
+    "oyster",
+    "oyster shell",
+    "package of fresh cut fruit",
+    "peach",
+    "pill",
+    "polka dot",
+    "prawn cracker",
+    "sausage",
+    "seagull",
+    "shallot",
+    "shirt",
+    "skateboard",
+    "toilet paper roll"
+  ]
+}

debug/config_args_raw.json

@@ -0,0 +1,30 @@
+{
+  "config_file": "config/cfg_fsc147_vit_b.py",
+  "options": {
+    "text_encoder_type": "checkpoints/bert-base-uncased"
+  },
+  "datasets": "config/datasets_fsc147.json",
+  "remove_difficult": false,
+  "fix_size": false,
+  "output_dir": "./debug",
+  "note": "",
+  "device": "cuda",
+  "seed": 42,
+  "resume": "",
+  "pretrain_model_path": "checkpoints/groundingdino_swinb_cogcoor.pth",
+  "finetune_ignore": null,
+  "start_epoch": 0,
+  "eval": false,
+  "num_workers": 8,
+  "test": false,
+  "debug": false,
+  "find_unused_params": false,
+  "save_results": false,
+  "save_log": false,
+  "world_size": 1,
+  "dist_url": "env://",
+  "rank": 0,
+  "local_rank": 0,
+  "amp": false,
+  "distributed": false
+}

debug/config_cfg.py

@@ -0,0 +1,140 @@
+data_aug_scales = [480, 512, 544, 576, 608, 640, 672, 704, 736, 768, 800]
+data_aug_max_size = 1333
+data_aug_scales2_resize = [400, 500, 600]
+data_aug_scales2_crop = [384, 600]
+data_aug_scale_overlap = None
+batch_size = 4
+modelname = 'groundingdino'
+backbone = 'swin_B_384_22k'
+position_embedding = 'sine'
+pe_temperatureH = 20
+pe_temperatureW = 20
+return_interm_indices = [1, 2, 3]
+enc_layers = 6
+dec_layers = 6
+pre_norm = False
+dim_feedforward = 2048
+hidden_dim = 256
+dropout = 0.0
+nheads = 8
+num_queries = 900
+query_dim = 4
+num_patterns = 0
+num_feature_levels = 4
+enc_n_points = 4
+dec_n_points = 4
+two_stage_type = 'standard'
+two_stage_bbox_embed_share = False
+two_stage_class_embed_share = False
+transformer_activation = 'relu'
+dec_pred_bbox_embed_share = True
+dn_box_noise_scale = 1.0
+dn_label_noise_ratio = 0.5
+dn_label_coef = 1.0
+dn_bbox_coef = 1.0
+embed_init_tgt = True
+dn_labelbook_size = 91
+max_text_len = 256
+text_encoder_type = 'checkpoints/bert-base-uncased'
+use_text_enhancer = True
+use_fusion_layer = True
+use_checkpoint = True
+use_transformer_ckpt = True
+use_text_cross_attention = True
+text_dropout = 0.0
+fusion_dropout = 0.0
+fusion_droppath = 0.1
+sub_sentence_present = True
+max_labels = 90
+lr = 0.0001
+backbone_freeze_keywords = None
+freeze_keywords = ['backbone.0', 'bert']
+lr_backbone = 1e-05
+lr_backbone_names = ['backbone.0', 'bert']
+lr_linear_proj_mult = 1e-05
+lr_linear_proj_names = ['ref_point_head', 'sampling_offsets']
+weight_decay = 0.0001
+param_dict_type = 'ddetr_in_mmdet'
+ddetr_lr_param = False
+epochs = 30
+lr_drop = 10
+save_checkpoint_interval = 10
+clip_max_norm = 0.1
+onecyclelr = False
+multi_step_lr = False
+lr_drop_list = [10, 20]
+frozen_weights = None
+dilation = False
+pdetr3_bbox_embed_diff_each_layer = False
+pdetr3_refHW = -1
+random_refpoints_xy = False
+fix_refpoints_hw = -1
+dabdetr_yolo_like_anchor_update = False
+dabdetr_deformable_encoder = False
+dabdetr_deformable_decoder = False
+use_deformable_box_attn = False
+box_attn_type = 'roi_align'
+dec_layer_number = None
+decoder_layer_noise = False
+dln_xy_noise = 0.2
+dln_hw_noise = 0.2
+add_channel_attention = False
+add_pos_value = False
+two_stage_pat_embed = 0
+two_stage_add_query_num = 0
+two_stage_learn_wh = False
+two_stage_default_hw = 0.05
+two_stage_keep_all_tokens = False
+num_select = 900
+batch_norm_type = 'FrozenBatchNorm2d'
+masks = False
+aux_loss = True
+set_cost_class = 5.0
+set_cost_bbox = 1.0
+set_cost_giou = 0.0
+cls_loss_coef = 5.0
+bbox_loss_coef = 1.0
+giou_loss_coef = 0.0
+enc_loss_coef = 1.0
+interm_loss_coef = 1.0
+no_interm_box_loss = False
+mask_loss_coef = 1.0
+dice_loss_coef = 1.0
+focal_alpha = 0.25
+focal_gamma = 2.0
+decoder_sa_type = 'sa'
+matcher_type = 'HungarianMatcher'
+decoder_module_seq = ['sa', 'ca', 'ffn']
+nms_iou_threshold = -1
+dec_pred_class_embed_share = True
+match_unstable_error = True
+use_detached_boxes_dec_out = False
+dn_scalar = 100
+box_threshold = 0.23
+text_threshold = 0
+use_coco_eval = False
+label_list = [
+    'alcohol bottle', 'baguette roll', 'ball', 'banana', 'bead', 'bee',
+    'birthday candle', 'biscuit', 'boat', 'bottle', 'bowl', 'box',
+    'bread roll', 'brick', 'buffalo', 'bun', 'calamari ring', 'can', 'candle',
+    'cap', 'car', 'cartridge', 'cassette', 'cement bag', 'cereal',
+    'chewing gum piece', 'chopstick', 'clam', 'coffee bean', 'coin',
+    'cotton ball', 'cow', 'crane', 'crayon', 'croissant', 'crow', 'cup',
+    'cupcake', 'cupcake holder', 'fish', 'gemstone', 'go game piece', 'goat',
+    'goldfish snack', 'goose', 'ice cream', 'ice cream cone', 'instant noodle',
+    'jade stone', 'jeans', 'kidney bean', 'kitchen towel', 'lighter',
+    'lipstick', 'm&m piece', 'macaron', 'match', 'meat skewer', 'mini blind',
+    'mosaic tile', 'naan bread', 'nail', 'nut', 'onion ring', 'orange',
+    'pearl', 'pen', 'pencil', 'penguin', 'pepper', 'person', 'pigeon', 'plate',
+    'polka dot tile', 'potato', 'rice bag', 'roof tile', 'screw', 'shoe',
+    'spoon', 'spring roll', 'stair', 'stapler pin', 'straw',
+    'supermarket shelf', 'swan', 'tomato', 'watermelon', 'window', 'zebra'
+]
+val_label_list = [
+    'ant', 'bird', 'book', 'bottle cap', 'bullet', 'camel', 'chair',
+    'chicken wing', 'donut', 'donut holder', 'flamingo', 'flower',
+    'flower pot', 'grape', 'horse', 'kiwi', 'milk carton', 'oyster',
+    'oyster shell', 'package of fresh cut fruit', 'peach', 'pill', 'polka dot',
+    'prawn cracker', 'sausage', 'seagull', 'shallot', 'shirt', 'skateboard',
+    'toilet paper roll'
+]

flagged/log.csv

@@ -0,0 +1,215 @@
+Input Image,Specify object to count here:,output,flag,username,timestamp
+flagged/Input Image/88e8b1b238c064836037/HVITa2016a_000011.JPG,bird,flagged/output/a686c9813d1a6c436907/image.webp,,,2024-06-09 09:33:54.235722
+flagged/Input Image/875b25d454777d421c42/221113_ECCC_MRADUL_ELEPHANT_MIA_RHEA_EVENING_WALK_KEEPER-8-1920x900.jpg,people and elephants,,,flagged/Detected Instances/76fafa9d17afaaf9eb38/image.webp,4,,,2024-06-18 16:54:16.644520
+flagged/Input Image/037d1ea9a55366ad2940/2.jpg,shell,,"{""image"": ""flagged/Specify object to count with visual exemplars here/fceed5d5292244ecdd73/2.jpg"", ""points"": [[380.0, 211.0, 2.0, 483.0, 317.0, 3.0]]}",flagged/Detected Instances/27c94b87a5ced00b9dd1/image.webp,8,,,2024-06-22 08:03:07.969202
+flagged/Input Image/b8faa04105f66c5a71d1/7.jpg,woman,,,flagged/Detected Instances/cd67c449c0a10a91f35b/image.webp,1,,,2024-06-22 08:08:57.657953
+flagged/Input Image/350e00785dada901c520/7.jpg,woman,,,flagged/Detected Instances/b16f6a58d39f8b72ca49/image.webp,1,,,2024-06-22 08:18:43.429051
+flagged/Input Image/12d92058a4ac08b10dea/7.jpg,woman,,"{""image"": ""flagged/Specify object to count with visual exemplars here/38245c2c066faff3cae8/7.jpg"", ""points"": [[113.0, 132.0, 2.0, 260.0, 315.0, 3.0]]}",flagged/Detected Instances/b941dacc2ffe391d2e16/image.webp,10,,,2024-06-22 08:19:03.379704
+flagged/Input Image/9b95bcefbe76b48d43ea/39.jpg,,,"{""image"": ""flagged/Specify object to count with visual exemplars here/f7cb8955646797bfe577/39.jpg"", ""points"": [[218.0, 225.0, 2.0, 295.0, 282.0, 3.0]]}",flagged/Detected Instances/f1930c50e7cf23e1bab8/image.webp,20,,,2024-06-22 08:19:37.258705
+flagged/Input Image/a29aa3c68b71af80ceac/39.jpg,tomato,,"{""image"": ""flagged/Specify object to count with visual exemplars here/80430ff258140b26bd0e/39.jpg"", ""points"": [[218.0, 225.0, 2.0, 295.0, 282.0, 3.0]]}",flagged/Detected Instances/fc4d452719e25259f01d/image.webp,19,,,2024-06-22 08:19:58.499357
+flagged/Input Image/bdf516de7a637158f274/HVITa2016a_000011.JPG,bird,,,flagged/Detected Instances/d0d2e5d0a4a0ac61ac4d/image.webp,30,,,2024-06-22 09:39:09.374430
+flagged/Input Image/9d6b2710e64f8889dd64/HVITa2016a_000011.JPG,,,"{""image"": ""flagged/Specify object to count with visual exemplars here/453f32945cd9e681b483/KAPWa2019a_000132.JPG"", ""points"": [[510.0, 150.0, 2.0, 534.0, 176.0, 3.0], [684.0, 205.0, 2.0, 712.0, 224.0, 3.0]]}",flagged/Detected Instances/5c1af4b6118751ff9b9d/image.webp,32,,,2024-06-22 09:40:38.259957
+flagged/Input Image/5ca9673ad1540cff6509/HVITa2016a_000011.JPG,bird,,"{""image"": ""flagged/Specify object to count with visual exemplars here/5df88f87ed6641277db1/KAPWa2019a_000132.JPG"", ""points"": [[744.0, 429.0, 2.0, 763.0, 461.0, 3.0]]}",flagged/Detected Instances/f44fe32f6932dd6574b3/image.webp,31,,,2024-06-22 09:41:55.748604
+flagged/Input Image/b6a179bab6c6189ddad7/55.jpg,tomato,,,flagged/Detected Instances/6d315017249048f38b52/image.webp,50,,,2024-06-22 10:41:12.653937
+flagged/Input Image/9bf0e79d190639880832/55.jpg,tomato,,"{""image"": ""flagged/Specify object to count with visual exemplars here/fe802eb93b73f4dae6b7/42.jpg"", ""points"": [[180.0, 128.0, 2.0, 224.0, 162.0, 3.0]]}",flagged/Detected Instances/f1b139f485f3c6049bb5/image.webp,50,,,2024-06-22 10:41:37.183988
+flagged/Input Image/80d1ea6bfc5bc8002c62/28.jpg,cinnamon roll,,,flagged/Detected Instances/caac9eeb8dcadfd732cb/image.webp,8,,,2024-06-22 11:13:59.554789
+flagged/Input Image/3f37e6da6e272866e8de/3.jpg,hot air balloon,,,flagged/Detected Instances/24844df4b3df513a0c36/image.webp,11,,,2024-06-22 14:11:32.948564
+flagged/Input Image/5e7dc59cf0a41b9bc5fb/fish images.jpeg,fish,,,flagged/Detected Instances/a96993ece7e66065ba7d/image.webp,77,,,2024-06-22 14:16:37.842263
+flagged/Input Image/57a992296c0ec254b864/Fish 2.jpeg,fish,,,flagged/Detected Instances/2c9d451701856fe3371d/image.webp,900,,,2024-06-22 14:20:50.590216
+flagged/Input Image/6c3c11094015679fa790/Fish 3.jpeg,fish,,,flagged/Detected Instances/4e75007f5ef68e4ca3d6/image.webp,839,,,2024-06-22 14:22:09.322830
+flagged/Input Image/31698b5d2863b6756bfa/Fish 4.jpeg,fish,,,flagged/Detected Instances/9d32c5fb797b4985cee0/image.webp,278,,,2024-06-22 14:25:47.760426
+flagged/Input Image/5ce8e5a0fa4008aec1f2/Fish 4.jpeg,fish,,"{""image"": ""flagged/Specify object to count with visual exemplars here/6780b6a4f82e7d10d5cf/fish images.jpeg"", ""points"": [[63.0, 65.0, 2.0, 478.0, 217.0, 3.0]]}",flagged/Detected Instances/c6b81dedf92ab0af4153/image.webp,286,,,2024-06-22 14:28:36.854956
+flagged/Input Image/00be0bde65646930f991/Fish 4.jpeg,fish,,"{""image"": ""flagged/Specify object to count with visual exemplars here/25a3e4831f706db4a521/Fish 3.jpeg"", ""points"": [[319.0, 221.0, 2.0, 650.0, 516.0, 3.0]]}",flagged/Detected Instances/25d123e67ec8644c5f02/image.webp,277,,,2024-06-22 14:29:09.194220
+flagged/Input Image/542e2646fed421a4d5e5/6f91f056-e447-4316-aae1-b1ca57b55ec6.jpeg,Man,,,flagged/Detected Instances/4e890d293a4cc098b2b5/image.webp,1,,,2024-06-22 14:29:55.868319
+flagged/Input Image/5571766c4ae584f798f6/Fish 3.jpeg,fish and fish with weird and large mouth,,,flagged/Detected Instances/9f53a1f8ee034afcb29f/image.webp,347,,,2024-06-22 14:32:29.009512
+flagged/Input Image/df8f6e4f3c2658165b32/Fish 3.jpeg,fish and fish with weird and large mouth,,,flagged/Detected Instances/9c453d75fc3f0b7ba779/image.webp,347,,,2024-06-22 14:33:38.795075
+flagged/Input Image/52226488b0c44d964af9/Fish 3.jpeg,All fishes big and small,,,flagged/Detected Instances/bf38e737886a0ac61746/image.webp,609,,,2024-06-22 14:34:34.754255
+flagged/Input Image/caae84890c10f63e5c53/Fish 3.jpeg,All fish big and small,,,flagged/Detected Instances/d209f5a62318a3cc9a83/image.webp,653,,,2024-06-22 14:34:49.612537
+flagged/Input Image/b1986e95b8a8df57d81b/Fish 3.jpeg,A big fish and all small fish,,,flagged/Detected Instances/10fdbef3f9c29b00d343/image.webp,655,,,2024-06-22 14:36:39.308451
+flagged/Input Image/b320c79d24f53c48cb36/Fish 3.jpeg,A big fish and all small fishes,,,flagged/Detected Instances/83948fd3cdfc64acba91/image.webp,663,,,2024-06-22 14:37:47.440375
+flagged/Input Image/abf54b47c8c97ae29e2d/Fish 3.jpeg,,,"{""image"": ""flagged/Specify object to count with visual exemplars here/0b10284a52f55d2b997c/Fish 3.jpeg"", ""points"": [[738.0, 295.0, 2.0, 917.0, 469.0, 3.0]]}",flagged/Detected Instances/2bb2c04de9c3c15284a2/image.webp,595,,,2024-06-22 14:42:47.880099
+flagged/Input Image/485ab0f7f21cef342c33/Fish 3.jpeg,,,"{""image"": ""flagged/Specify object to count with visual exemplars here/a2645b60d75a35644c30/Fish 3.jpeg"", ""points"": [[717.0, 65.0, 2.0, 934.0, 232.0, 3.0], [812.0, 55.0, 2.0, 812.0, 155.0, 3.0]]}",flagged/Detected Instances/bb94b03bbae3744cb687/image.webp,710,,,2024-06-22 14:43:57.785652
+flagged/Input Image/eaf2d446d413a9b288cf/Fish 3.jpeg,A big fish and all small fishes,,"{""image"": ""flagged/Specify object to count with visual exemplars here/8848ccbc7b6f9793d87a/Fish 3.jpeg"", ""points"": [[717.0, 65.0, 2.0, 934.0, 232.0, 3.0], [812.0, 55.0, 2.0, 812.0, 155.0, 3.0]]}",flagged/Detected Instances/31643c3688da9b4fa531/image.webp,716,,,2024-06-22 14:44:56.566214
+flagged/Input Image/432fc52db57386081a7f/Fish 3.jpeg,A big fish and all small fishes,,"{""image"": ""flagged/Specify object to count with visual exemplars here/b3b4cf1ebceac265a27a/Fish 3.jpeg"", ""points"": [[202.0, 10.0, 2.0, 678.0, 280.0, 3.0]]}",flagged/Detected Instances/596768cb96359baae656/image.webp,736,,,2024-06-22 14:46:10.694967
+flagged/Input Image/a12f368ae7c4c5fd3da6/Fish 3.jpeg,,,"{""image"": ""flagged/Specify object to count with visual exemplars here/a59d14a310592e735186/Fish 3.jpeg"", ""points"": [[44.0, 96.0, 2.0, 83.0, 148.0, 3.0]]}",flagged/Detected Instances/663747d5d817f3b6f17b/image.webp,603,,,2024-06-22 14:50:04.561496
+flagged/Input Image/e081cd4b3f8c0a233e93/HVITa2016a_000011.JPG,bird,,,flagged/Detected Instances/ef9c94eaef8c71a59be7/image.webp,30,,,2024-06-22 14:59:44.195345
+flagged/Input Image/037b6400419594ee083f/2024-06-20 21.21.07.jpg,apples,,,flagged/Detected Instances/69304bef5b817c19a40d/image.webp,0,,,2024-06-22 21:16:37.304942
+flagged/Input Image/7fb25053e99ea97dec5d/IMG_4345.jpeg,bottle of water,,,flagged/Detected Instances/cd9ef90fd76f3c5bc28d/image.webp,5,,,2024-06-22 21:18:02.687111
+flagged/Input Image/ac22a60546af5b9e27d5/Julie_Chandelier_South-Georgia_Gold-Harbour_King-Penguins-colony-2-scaled.jpg,,,"{""image"": ""flagged/Specify object to count with visual exemplars here/9df284bf10e2e549b58b/900.jpeg"", ""points"": [[254.0, 43.0, 2.0, 615.0, 646.0, 3.0]]}",flagged/Detected Instances/6afde69e615172163cd9/image.webp,19,,,2024-06-22 21:24:24.843762
+flagged/Input Image/1e3fbaf70a0e480fcc7e/photo-1536842717890-4804617eb992.jpeg,balloon,,,flagged/Detected Instances/9e35897bb2b104feaf01/image.webp,21,,,2024-06-23 18:19:01.914350
+flagged/Input Image/8c8253a9745752f7ac5c/photo-1533551131698-c87deddfa2d6.jpeg,person,,,flagged/Detected Instances/f73654df476be2e2025a/image.webp,392,,,2024-06-23 18:20:38.923697
+flagged/Input Image/ba84aaaf18399d9d7285/photo-1533551131698-c87deddfa2d6.jpeg,person,,"{""image"": ""flagged/Specify object to count with visual exemplars here/2887c60fea7795aacecb/photo-1533551131698-c87deddfa2d6.jpeg"", ""points"": [[471.0, 235.0, 2.0, 529.0, 335.0, 3.0]]}",flagged/Detected Instances/5128711a7e3b3b296705/image.webp,379,,,2024-06-23 18:21:23.083042
+flagged/Input Image/288c3bba8e73832aa8a2/photo-1533551131698-c87deddfa2d6.jpeg,,,"{""image"": ""flagged/Specify object to count with visual exemplars here/8af1f367a37b6758f689/photo-1533551131698-c87deddfa2d6.jpeg"", ""points"": [[471.0, 235.0, 2.0, 529.0, 335.0, 3.0]]}",flagged/Detected Instances/26a1963fd9197c1e4c1d/image.webp,312,,,2024-06-23 18:21:38.559234
+flagged/Input Image/254816dc70cdd88ffa6c/photo-1533551131698-c87deddfa2d6.jpeg,person,,,flagged/Detected Instances/6bb3d26c43dc6b20eeba/image.webp,392,,,2024-06-23 18:21:50.311989
+flagged/Input Image/25d39fcb0445819504c4/photo-1533551131698-c87deddfa2d6.jpeg,person,,"{""image"": ""flagged/Specify object to count with visual exemplars here/005d56113a6597d456d6/photo-1533551131698-c87deddfa2d6.jpeg"", ""points"": [[473.0, 242.0, 2.0, 526.0, 339.0, 3.0]]}",flagged/Detected Instances/ef8e0debab490fee3637/image.webp,381,,,2024-06-23 18:22:20.986792
+flagged/Input Image/2bc6a37e212dcc397a14/JUTeHjN.jpg,,,"{""image"": ""flagged/Specify object to count with visual exemplars here/1dfc2902049a063b4d3b/JUTeHjN.jpg"", ""points"": [[274.0, 559.0, 2.0, 301.0, 639.0, 3.0]]}",flagged/Detected Instances/d0472660028ee835a909/image.webp,591,,,2024-06-23 18:26:11.912055
+flagged/Input Image/cc6b05c49c410ce714a4/JUTeHjN.jpg,waldo,,"{""image"": ""flagged/Specify object to count with visual exemplars here/0354019aa33aa8cc1b4c/JUTeHjN.jpg"", ""points"": [[274.0, 559.0, 2.0, 301.0, 639.0, 3.0]]}",flagged/Detected Instances/e7d80138cd030c1a4de7/image.webp,707,,,2024-06-23 18:26:56.983776
+flagged/Input Image/c1f5e295668daa58b78b/La_horde_-_Surfers_riding_a_wave_in_Paea_Tahiti.jpg,people,,,flagged/Detected Instances/2b8161f3e7ac96bc81a6/image.webp,6,,,2024-06-24 08:54:20.028204
+flagged/Input Image/3077b7ba647265981dac/Screenshot from 2024-06-18 07-24-03.png,letter t,,,flagged/Detected Instances/d0f7006b69b870c7c227/image.webp,60,,,2024-06-24 13:39:18.019532
+flagged/Input Image/9a11e4fe8738ffdc2176/Screenshot from 2024-06-18 07-24-03.png,letter t,,"{""image"": ""flagged/Specify object to count with visual exemplars here/441e53d9531d94725262/hello.png"", ""points"": []}",flagged/Detected Instances/0465e358fdb89e04a940/image.webp,60,,,2024-06-24 13:40:57.783042
+flagged/Input Image/afa9d75359a32a6d4b41/Screenshot from 2024-06-18 07-24-03.png,letter t,,"{""image"": ""flagged/Specify object to count with visual exemplars here/037cdd9bcd4683726fc5/hello.png"", ""points"": []}",flagged/Detected Instances/2f592103e85d7e6e527a/image.webp,60,,,2024-06-24 13:41:29.438433
+flagged/Input Image/57809c4a9bafbbd30918/Screenshot from 2024-06-18 07-24-03.png,Ts,,"{""image"": ""flagged/Specify object to count with visual exemplars here/e22fa571644aa160d86c/hello.png"", ""points"": []}",flagged/Detected Instances/c318b58b8c88b7710e0c/image.webp,6,,,2024-06-24 13:41:45.916615
+flagged/Input Image/c9412ee3b51eacd33f52/06.jpg_instances.jpg,fish,,,flagged/Detected Instances/a6c2ad6ab9379e0211fa/image.webp,9,,,2024-06-24 15:07:10.854678
+flagged/Input Image/e2976e4738abc316313f/word-image-233.jpeg,leaves,,,flagged/Detected Instances/e8eab3eeef0d444bbb1a/image.webp,10,,,2024-06-24 15:53:21.817485
+flagged/Input Image/be3ab6fa387c0a0194f4/word-image-233.jpeg,leaf,,,flagged/Detected Instances/90dc5680d2ec80c0eac8/image.webp,10,,,2024-06-24 15:55:43.676702
+flagged/Input Image/8ee3a4ce6a33884d03ae/2.jpg,sea shell,,,flagged/Detected Instances/f0d8fd3020ac899200f2/image.webp,8,,,2024-06-24 15:57:17.391770
+flagged/Input Image/74980bc345053ad14f89/215.jpg,grape,,,flagged/Detected Instances/e73eae638535e8770764/image.webp,243,,,2024-06-24 15:58:37.852766
+flagged/Input Image/d206f8d836baf220ac41/HVITa2016a_000011.JPG,bird,,,flagged/Detected Instances/9949e21c2f5bacf777a1/image.webp,30,,,2024-06-24 16:27:43.463142
+flagged/Input Image/86a081dc917ef9d837c7/HVITa2016a_000011.JPG,,,"{""image"": ""flagged/Specify object to count with visual exemplars here/64abdac51a65819724a6/HVITa2016a_000011.JPG"", ""points"": [[1230.0, 118.0, 2.0, 1292.0, 201.0, 3.0], [830.0, 550.0, 2.0, 906.0, 623.0, 3.0]]}",flagged/Detected Instances/d92c3d3991cfcefde001/image.webp,33,,,2024-06-24 16:28:22.627606
+flagged/Input Image/35151af517a911bcc223/HVITa2016a_000011.JPG,bird,,"{""image"": ""flagged/Specify object to count with visual exemplars here/4f611be7fe99047414ae/HVITa2016a_000011.JPG"", ""points"": [[1230.0, 118.0, 2.0, 1292.0, 201.0, 3.0], [830.0, 550.0, 2.0, 906.0, 623.0, 3.0]]}",flagged/Detected Instances/10125734c7afa38ed087/image.webp,31,,,2024-06-24 16:28:36.193649
+flagged/Input Image/bba1422fdfed279a73af/HVITa2016a_000011.JPG,bird,,flagged/Detected Instances/bec28a276df78afe343d/image.webp,30,,,2024-06-24 16:41:23.500662
+flagged/Input Image/d7762005e6674cf12e5c/HVITa2016a_000011.JPG,bird,,flagged/Detected Instances/cca9e646a051b17f95f5/image.webp,30,,,2024-06-24 17:55:54.474905
+flagged/Input Image/81aa794f290cedf734c3/HVITa2016a_000011.JPG,,"{""image"": ""flagged/Specify object to count with visual exemplars here/6da159a3958dedc078b2/HVITa2016a_000011.JPG"", ""points"": []}",flagged/Detected Instances/02faafb810646f2ea9b4/image.webp,0,,,2024-06-24 17:56:45.209436
+flagged/Input Image/7b8f501cc28c308ff8b6/HVITa2016a_000011.JPG,,"{""image"": ""flagged/Specify object to count with visual exemplars here/4a30c21797e2739e4abc/HVITa2016a_000011.JPG"", ""points"": [[943.0, 374.0, 2.0, 1057.0, 485.0, 3.0]]}",flagged/Detected Instances/e27a207c035db1ab701e/image.webp,32,,,2024-06-24 18:03:25.307742
+flagged/Input Image/7af486902d9e28508494/HVITa2016a_000011.JPG,,"{""image"": ""flagged/Specify object to count with visual exemplars here/4a1a53289daf6c4ab771/HVITa2016a_000011.JPG"", ""points"": []}",flagged/Detected Instances/2d8b8e39fc9512762815/image.webp,0,,,2024-06-24 18:43:48.203081
+flagged/Input Image/55afacf49af9b719cdda/HVITa2016a_000011.JPG,bird,"{""image"": ""flagged/Specify object to count with visual exemplars here/30839117a5e1c2687952/HVITa2016a_000011.JPG"", ""points"": [[1230.0, 118.0, 2.0, 1292.0, 201.0, 3.0], [830.0, 550.0, 2.0, 906.0, 623.0, 3.0]]}",flagged/Detected Instances/c9d49f2d15a7de44e80d/image.webp,31,,,2024-06-24 18:52:25.941469
+flagged/Input Image/266f273a429e2a835b60/HVITa2016a_000011.JPG,,"{""image"": ""flagged/Specify object to count with visual exemplars here/ea17095cd854670fd8db/HVITa2016a_000011.JPG"", ""points"": []}",flagged/Detected Instances/7f02a76d21f7e077559e/image.webp,0,,,2024-06-25 08:53:21.768814
+flagged/Input Image/e93ffc35e1c7c113beec/HVITa2016a_000011.JPG,,"{""image"": ""flagged/Specify object to count with visual exemplars here/af5231d7685032d2c2fe/HVITa2016a_000011.JPG"", ""points"": []}",flagged/Detected Instances/5cd4eb530194572ba703/image.webp,0,,,2024-06-25 09:16:55.226689
+flagged/Input Image/733b82dbcccd8aae5c53/HVITa2016a_000011.JPG,,"{""image"": ""flagged/Specify object to count with visual exemplars here/be9a0871ad384d450dbc/HVITa2016a_000011.JPG"", ""points"": []}",flagged/Detected Instances/d9e693ae5adb9314a442/image.webp,0,,,2024-06-25 09:22:48.487486
+flagged/Input Image/180e288224fa47f7e40e/HVITa2016a_000011.JPG,,"{""image"": ""flagged/Specify object to count with visual exemplars here/3b47f1cbc62a52b89a35/HVITa2016a_000011.JPG"", ""points"": []}",flagged/Detected Instances/c06ad461ce25c9e93519/image.webp,0,,,2024-06-25 09:37:10.918912
+flagged/Input Image/f6c46243c9e41736a91b/HVITa2016a_000011.JPG,bird,,flagged/Detected Instances/f834fe8aa737c6602250/image.webp,30,,,2024-06-25 09:40:18.398469
+flagged/Input Image/02e6dd294378714d2c60/HVITa2016a_000011.JPG,,"{""image"": ""flagged/Specify object to count with visual exemplars here/7a28c64f2d7f386ba90f/HVITa2016a_000011.JPG"", ""points"": []}",flagged/Detected Instances/58939acce9372835d94f/image.webp,0,,,2024-06-25 09:40:26.315283
+flagged/Input Image/ef125c4204764412c3f7/HVITa2016a_000011.JPG,,"{""image"": ""flagged/Specify object to count with visual exemplars here/5130a669c02f296bb278/HVITa2016a_000011.JPG"", ""points"": []}",flagged/Detected Instances/befab1125336c5e50181/image.webp,0,,,2024-06-25 10:05:42.679549
+flagged/Input Image/feae1a999efc1b161e4a/HVITa2016a_000011.JPG,,"{""image"": ""flagged/Specify object to count with visual exemplars here/d8f9a71db6b70a468a0f/HVITa2016a_000011.JPG"", ""points"": []}",flagged/Detected Instances/ec35f973d14b7fc08792/image.webp,0,,,2024-06-25 10:14:45.281246
+flagged/Input Image/b721bed7a52d56eabff8/HVITa2016a_000011.JPG,,"{""image"": ""flagged/Specify object to count with visual exemplars here/8df3b7b4105c8bbed352/HVITa2016a_000011.JPG"", ""points"": [[1224.0, 99.0, 2.0, 1308.0, 215.0, 3.0]]}",flagged/Detected Instances/1ab68fe2b72b062794c3/image.webp,32,,,2024-06-25 10:15:02.226475
+flagged/Input Image/30c0dd916e13e8cdf40a/HVITa2016a_000011.JPG,,"{""image"": ""flagged/Specify object to count with visual exemplars here/3183ccd48e9876f48116/HVITa2016a_000011.JPG"", ""points"": []}",flagged/Detected Instances/486ddeee601a5658d3c1/image.webp,0,,,2024-06-25 10:17:04.340172
+flagged/Input Image/59fcdf686244d693b758/HVITa2016a_000011.JPG,bird,,flagged/Detected Instances/1c4afcc87d7b113064e0/image.webp,30,,,2024-06-25 10:25:55.780204
+flagged/Input Image/0faa5dfd6e852c50d8ff/HVITa2016a_000011.JPG,,"{""image"": ""flagged/Specify object to count with visual exemplars here/5e3aabee37bf681d51a4/HVITa2016a_000011.JPG"", ""points"": []}",flagged/Detected Instances/2cd238061edc54130cdf/image.webp,0,,,2024-06-25 10:26:05.084390
+flagged/Input Image/3e839a4e8ebe95002a20/HVITa2016a_000011.JPG,bird,"{""image"": ""flagged/Specify object to count with visual exemplars here/f0cd0be5ea8ccbb26d0f/HVITa2016a_000011.JPG"", ""points"": []}",flagged/Detected Instances/b8c1235f9f45df306ef9/image.webp,30,,,2024-06-25 10:26:13.747497
+flagged/Input Image/aea44fcfb41612489dad/HVITa2016a_000011.JPG,,"{""image"": ""flagged/Specify object to count with visual exemplars here/da461ef7d215c55a5e7c/HVITa2016a_000011.JPG"", ""points"": []}",flagged/Detected Instances/cf0299a8a5c332b5c02f/image.webp,0,,,2024-06-25 10:27:17.764060
+flagged/Input Image/defc48612eca2926a7b7/bird-1.JPG,bird,"{""image"": ""flagged/Specify object to count with visual exemplars here/bf6d9e9b6fd1b4962765/bird-2.JPG"", ""points"": []}",flagged/Detected Instances/8a857874d6e33125d752/image.webp,30,,,2024-06-25 10:56:25.591931
+flagged/Input Image/02215a04cfb823dd7796/bird-1.JPG,bird,"{""image"": ""flagged/Specify object to count with visual exemplars here/54a44757b38db813d618/bird-2.JPG"", ""points"": [[608.0, 232.0, 2.0, 640.0, 263.0, 3.0]]}",flagged/Detected Instances/c0ed66b9cad56b199aa2/image.webp,31,,,2024-06-25 10:59:54.183533
+flagged/Input Image/df6cea1edf80683a3767/bird-1.JPG,bird,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/0dce00cba984dd532757/bird-2.JPG"", ""points"": [[503.0, 142.0, 2.0, 542.0, 184.0, 3.0]]}",flagged/Detected Instances/c1f230dd1e883d9e32e7/image.webp,31,,,2024-06-25 11:15:22.282716
+flagged/Input Image/8d0fb36ef83e86fd525a/strawberry.jpg,strawberry,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/e646d85a66c06474d22d/strawberry.jpg"", ""points"": [[160.0, 68.0, 2.0, 197.0, 138.0, 3.0]]}",flagged/Detected Instances/ed9419605b016fba2dc5/image.webp,16,,,2024-06-25 11:30:30.236817
+flagged/Input Image/b49396917cf5ecc697fa/strawberry.jpg,blueberry,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/f3fd414caafcebcbda4a/strawberry.jpg"", ""points"": []}",flagged/Detected Instances/8632f9c1bb05eee15dcc/image.webp,137,,,2024-06-25 11:30:43.620161
+flagged/Input Image/3dd1700ef3bcbe5657f7/strawberry.jpg,blueberry,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/5dfc0e9e8efa0af0c6a8/strawberry.jpg"", ""points"": [[253.0, 130.0, 2.0, 273.0, 152.0, 3.0]]}",flagged/Detected Instances/94ce91cd91c43e18974d/image.webp,133,,,2024-06-25 11:30:53.267942
+flagged/Input Image/cec4554c9b13aa4d346d/clipboard.png,fish,,flagged/Detected Instances/1830def027aa0d3f3d91/image.webp,434,,,2024-06-25 11:38:27.652346
+flagged/Input Image/032c2e1714788159e262/clipboard.png,fish,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/04a1a12a302c9c1db814/clipboard.png"", ""points"": [[1049.0, 213.0, 2.0, 1125.0, 238.0, 3.0]]}",flagged/Detected Instances/122ff6e9f6ff79c3e5a7/image.webp,411,,,2024-06-25 11:39:25.658141
+flagged/Input Image/5422ce8f8f4364819aaa/clipboard.png,fish,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/c32361eb60578a2bd4ac/clipboard.png"", ""points"": []}",flagged/Detected Instances/5d5df763986fec98e93f/image.webp,900,,,2024-06-25 11:41:22.135329
+flagged/Input Image/5072a86417c33eb593f3/clipboard.png,fish,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/028b7ae649c4dd8b7178/clipboard.png"", ""points"": [[1832.0, 158.0, 2.0, 1878.0, 223.0, 3.0]]}",flagged/Detected Instances/c65818f6571e4d168e43/image.webp,900,,,2024-06-25 11:41:46.118310
+flagged/Input Image/df7f9cdf52e79013f8cb/clipboard.png,fish,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/09eb9b77eff2e836b6ff/clipboard.png"", ""points"": []}",flagged/Detected Instances/b4a6138c92f1c746ccdb/image.webp,248,,,2024-06-25 11:42:43.425077
+flagged/Input Image/a9ccd350937abcb63291/clipboard.png,fish,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/19ce574f4470218dc7e0/clipboard.png"", ""points"": [[367.0, 183.0, 2.0, 451.0, 219.0, 3.0]]}",flagged/Detected Instances/d48cc5c8133c9068a0dd/image.webp,247,,,2024-06-25 11:43:33.822691
+flagged/Input Image/a3d2036f9c1ac12e961c/fish.jpg,fish,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/0c6da89b80bb001aa49e/fish.jpg"", ""points"": []}",flagged/Detected Instances/f27dfe196a1c41a69503/image.webp,248,,,2024-06-25 11:45:59.667866
+flagged/Input Image/46f4410a1aa98f5ee2af/strawberry.jpg,blueberry,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/2fd7f3ce369404120141/strawberry.jpg"", ""points"": []}",flagged/Detected Instances/0197e9eeac0554007c3f/image.webp,137,,,2024-06-25 11:46:32.895500
+flagged/Input Image/bb268ffa2de20ebe1c26/strawberry.jpg,blueberry,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/c3b2d652d852a7eab171/strawberry.jpg"", ""points"": [[342.0, 126.0, 2.0, 359.0, 144.0, 3.0]]}",flagged/Detected Instances/44f0249ecddfb4155756/image.webp,137,,,2024-06-25 11:46:42.177472
+flagged/Input Image/6c66434da1d48433ecb3/strawberry.jpg,blueberry,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/4d4cffebe4515feff0ec/strawberry.jpg"", ""points"": []}",flagged/Detected Instances/ab963c869b4f8d820e49/image.webp,137,,,2024-06-25 11:46:57.009420
+flagged/Input Image/a6875dccb70b93acd740/strawberry.jpg,blueberry,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/f481446bb34bcb55993f/strawberry.jpg"", ""points"": [[333.0, 167.0, 2.0, 353.0, 187.0, 3.0], [301.0, 100.0, 2.0, 321.0, 119.0, 3.0]]}",flagged/Detected Instances/bdf5762c87a5fb86e0b2/image.webp,137,,,2024-06-25 11:47:18.226311
+flagged/Input Image/44cef4325bfc0b2aa091/strawberry.jpg,strawberry,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/bcaa94ade9f2b6274f1c/strawberry.jpg"", ""points"": []}",flagged/Detected Instances/613e0b5dca0804e918c7/image.webp,16,,,2024-06-25 11:47:34.461816
+flagged/Input Image/6d93ce821c043cc1fa50/strawberry.jpg,blueberry,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/ae226a70b29c89cc4eb8/strawberry.jpg"", ""points"": []}",flagged/Detected Instances/00176ba3de1e4fbd2c79/image.webp,137,,,2024-06-25 11:47:45.327559
+flagged/Input Image/214729be550d569c1c68/bird-1.JPG,bird,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/f6eed5c264264d847ff6/bird-2.JPG"", ""points"": []}",flagged/Detected Instances/c12f210c5902dfd03e70/image.webp,30,,,2024-06-25 11:48:34.856890
+flagged/Input Image/8bf133a0564fd92cc890/balloon.jpg,hot air balloon,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/978150756fa4f1ef5027/balloon.jpg"", ""points"": []}",flagged/Detected Instances/5b47d64211343b63152a/image.webp,108,,,2024-06-25 12:25:57.589269
+flagged/Input Image/6429656af3e526d1c231/women.jpg,women,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/25350112732936ba546b/women.jpg"", ""points"": []}",flagged/Detected Instances/be57da4ba536ae483835/image.webp,5,,,2024-06-25 12:26:41.171433
+flagged/Input Image/c58a77f0b28ca5d81b43/women.jpg,men,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/130e25a655d338fdfe3e/women.jpg"", ""points"": []}",flagged/Detected Instances/e7e68bb358caa2a02fad/image.webp,6,,,2024-06-25 12:26:49.904389
+flagged/Input Image/a7f9b2e70fc2dd48ea61/women.jpg,man,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/cd6645e130252cb0ce54/women.jpg"", ""points"": []}",flagged/Detected Instances/7ffc5ec4d2638a7d5993/image.webp,4,,,2024-06-25 12:26:59.917970
+flagged/Input Image/42851096a5465744aff8/women.jpg,boy,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/fdaea5a42ce12dafba02/women.jpg"", ""points"": []}",flagged/Detected Instances/47cfc5281c09ccad9f3c/image.webp,1,,,2024-06-25 12:27:25.852703
+flagged/Input Image/67132ca41a6c48bc4ae8/women.jpg,girl,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/2410b57b1dae0d986036/women.jpg"", ""points"": []}",flagged/Detected Instances/b8ef07f2a0d78188051d/image.webp,5,,,2024-06-25 12:27:35.919728
+flagged/Input Image/d66d3772b1409e1852d8/strawberry.jpg,strawberry,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/2f7a42ec53940a767ab1/strawberry.jpg"", ""points"": []}",flagged/Detected Instances/db7ecc53cc8f29566958/image.webp,16,,,2024-06-25 12:29:35.376726
+flagged/Input Image/7493d467be9cfe1dff7c/strawberry.jpg,strawberry,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/8cab0399346d70c00fc0/strawberry.jpg"", ""points"": [[338.0, 123.0, 2.0, 361.0, 146.0, 3.0]]}",flagged/Detected Instances/f01f6a02720e21ce0150/image.webp,50,,,2024-06-25 12:29:45.148685
+flagged/Input Image/440c504736080f8142ad/strawberry.jpg,blueberry,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/92fd535788251da6ba09/strawberry.jpg"", ""points"": [[338.0, 123.0, 2.0, 361.0, 146.0, 3.0]]}",flagged/Detected Instances/990386fa1a9553c882db/image.webp,136,,,2024-06-25 12:29:57.059622
+flagged/Input Image/fed431554a261ab266a4/strawberry.jpg,blueberry,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/60afc4b94764dc018370/strawberry.jpg"", ""points"": []}",flagged/Detected Instances/5f048bdb5003eb64e17b/image.webp,137,,,2024-06-25 12:30:08.506742
+flagged/Input Image/5e49dbc792a65c4dbbb5/strawberry.jpg,blueberry,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/8415524ab37ceaf7469b/strawberry.jpg"", ""points"": [[338.0, 123.0, 2.0, 361.0, 146.0, 3.0], [257.0, 112.0, 2.0, 287.0, 136.0, 3.0]]}",flagged/Detected Instances/e56fa7694cc83977e48e/image.webp,137,,,2024-06-25 12:30:18.906628
+flagged/Input Image/daa4ac7838796afe20fd/bird-1.JPG,bird,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/69dcd045d1a47c1a45cc/bird-2.JPG"", ""points"": []}",flagged/Detected Instances/7c51fda26eae4f4660a6/image.webp,30,,,2024-06-25 12:30:29.347865
+flagged/Input Image/7ba4ec104b5a3a146474/bird-1.JPG,bird,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/164da9e8be790ea4ef79/bird-2.JPG"", ""points"": [[505.0, 147.0, 2.0, 536.0, 183.0, 3.0]]}",flagged/Detected Instances/f16c10ef2bc36b5dc69d/image.webp,31,,,2024-06-25 12:30:39.178460
+flagged/Input Image/420750e47befc70801e2/fish.jpg,fish,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/9e22ac918784790aa880/fish.jpg"", ""points"": []}",flagged/Detected Instances/27ff780f6a23afc048b2/image.webp,248,,,2024-06-25 12:31:01.375887
+flagged/Input Image/1feeba99c7d2efe9096e/women.jpg,girl,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/4f7337051dee8ff77653/women.jpg"", ""points"": []}",flagged/Detected Instances/ce0dd6ff795456176eab/image.webp,5,,,2024-06-25 12:31:17.643196
+flagged/Input Image/9f1c5da2733fa8c58ae8/women.jpg,boy,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/d41a6c081af41e9a2f62/women.jpg"", ""points"": []}",flagged/Detected Instances/620e679c66d0f9574fad/image.webp,1,,,2024-06-25 12:31:28.806538
+flagged/Input Image/30598862ab850cd6db31/balloon.jpg,hot air balloon,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/ff94594958fb2da48621/balloon.jpg"", ""points"": []}",flagged/Detected Instances/16993819652dc9bce0b8/image.webp,108,,,2024-06-25 12:31:44.635515
+flagged/Input Image/f9dbff9baf36cf6c8a91/balloon.jpg,hot air balloon,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/39ac7da130be6fb0cba1/balloon.jpg"", ""points"": [[255.0, 38.0, 2.0, 274.0, 60.0, 3.0]]}",flagged/Detected Instances/fbaa66b107151c86b5f3/image.webp,108,,,2024-06-25 12:31:53.008190
+flagged/Input Image/86ef5e44d86d09969fa3/deer.jpg,deer,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/fbc70e37fee443170ff4/deer.jpg"", ""points"": []}",flagged/Detected Instances/d1ab59ab9d6d21e0c4b3/image.webp,21,,,2024-06-25 12:32:16.020013
+flagged/Input Image/279428723858139b5a4c/apple.jpg,apple,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/ea6e4f53ba1edf38def6/apple.jpg"", ""points"": []}",flagged/Detected Instances/f45dc48dfdc3f7df6249/image.webp,568,,,2024-06-25 12:32:45.752592
+flagged/Input Image/8213864ae0272f59994f/apple.jpg,apple,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/d5d645dce3a0bedcf3c7/apple.jpg"", ""points"": [[316.0, 230.0, 2.0, 365.0, 273.0, 3.0]]}",flagged/Detected Instances/8f1288ccce0a4abba02c/image.webp,550,,,2024-06-25 12:33:07.817865
+flagged/Input Image/1482d3171ccab0738d15/egg.jpg,egg,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/937d3cd3614befcd3692/egg.jpg"", ""points"": []}",flagged/Detected Instances/577b9c0ccd91d4d879ed/image.webp,18,,,2024-06-25 12:33:27.798499
+flagged/Input Image/25f50a2c8a3aa7d190bd/egg.jpg,egg,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/d345ceb5293859fabf1c/egg.jpg"", ""points"": [[418.0, 85.0, 2.0, 472.0, 148.0, 3.0]]}",flagged/Detected Instances/87eb5f1e0f18dc6739f6/image.webp,18,,,2024-06-25 12:33:35.566274
+flagged/Input Image/9f0db0ff082f6b31f825/stamp.jpg,stamp,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/701c11e28665c59565cc/stamp.jpg"", ""points"": []}",flagged/Detected Instances/82cabb88e68e8ecc63ba/image.webp,45,,,2024-06-25 12:33:47.051392
+flagged/Input Image/fcb41365c9ac544c8898/green-pea.jpg,green pea,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/2b2c58e68049e4dade9d/green-pea.jpg"", ""points"": []}",flagged/Detected Instances/3f41f7d66283c8f516ec/image.webp,213,,,2024-06-25 12:33:59.684009
+flagged/Input Image/8a701be26fb9fb2d40e8/lego.jpg,lego,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/87e17ecb546906d99293/lego.jpg"", ""points"": []}",flagged/Detected Instances/e1adec40dc3e0880d984/image.webp,23,,,2024-06-25 12:34:29.213218
+flagged/Input Image/4673ed3f053269ba11d4/lego.jpg,lego,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/aae11dd32545f4f6e48a/lego.jpg"", ""points"": [[275.0, 166.0, 2.0, 294.0, 186.0, 3.0]]}",flagged/Detected Instances/0ed9da189663a3352717/image.webp,23,,,2024-06-25 12:34:41.379472
+flagged/Input Image/d956a7af5032db84ebf4/lego.jpg,lego,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/65d7ec1fb0e4f2a6b378/lego.jpg"", ""points"": [[275.0, 166.0, 2.0, 294.0, 186.0, 3.0], [269.0, 256.0, 2.0, 284.0, 273.0, 3.0]]}",flagged/Detected Instances/f0675e3f8e5fbae79512/image.webp,23,,,2024-06-25 12:34:50.644433
+flagged/Input Image/4a629b96d8745c48c739/lego.jpg,lego,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/a71c7ef0952943bac3d3/lego.jpg"", ""points"": [[275.0, 166.0, 2.0, 294.0, 186.0, 3.0], [269.0, 256.0, 2.0, 284.0, 273.0, 3.0], [124.0, 213.0, 2.0, 143.0, 232.0, 3.0]]}",flagged/Detected Instances/28d4f0037a145a77e53f/image.webp,23,,,2024-06-25 12:35:00.163591
+flagged/Input Image/be96be6a6471be11367b/fish.jpg,fish,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/e94ed1719ff8f3280d11/fish.jpg"", ""points"": []}",flagged/Detected Instances/4ad9cd81c3a99b08a811/image.webp,248,,,2024-06-25 12:37:35.824495
+flagged/Input Image/5c18d75a43fc788202de/egg.jpg,egg,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/1158914112f891b5efd0/egg.jpg"", ""points"": []}",flagged/Detected Instances/182a0d4d6453ae657719/image.webp,18,,,2024-06-25 12:38:00.758125
+flagged/Input Image/e0811fca1751db63b380/egg.jpg,,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/382cdaec8615c7e2fecb/egg.jpg"", ""points"": [[199.0, 86.0, 2.0, 248.0, 142.0, 3.0]]}",flagged/Detected Instances/2079378c09fdd2a9c7ce/image.webp,18,,,2024-06-25 12:38:18.291809
+flagged/Input Image/d2e387b8a61c19868cba/apple.jpg,apple,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/3ec031a562ec058e7731/apple.jpg"", ""points"": []}",flagged/Detected Instances/adf208e2ecead3e6a1f4/image.webp,568,,,2024-06-25 12:38:34.859201
+flagged/Input Image/5f63a040416264d96fc7/green-pea.jpg,green pea,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/517ad35ce8c0e44b6782/green-pea.jpg"", ""points"": []}",flagged/Detected Instances/66a8dfe7bfefebd3121d/image.webp,213,,,2024-06-25 12:39:07.587414
+flagged/Input Image/aae210f02dbbc2bdd515/stamp.jpg,stamp,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/422d122996b90782f0a3/stamp.jpg"", ""points"": []}",flagged/Detected Instances/6cbaa1c841d2eb3ec14d/image.webp,45,,,2024-06-25 12:42:10.960744
+flagged/Input Image/81719deb5b7c5097ebb0/stamp.jpg,square,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/959e12c05af1e92d9355/stamp.jpg"", ""points"": []}",flagged/Detected Instances/21ac1219abab8e1a1e3a/image.webp,45,,,2024-06-25 12:42:30.712204
+flagged/Input Image/41061285b258efc14011/stamp.jpg,25,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/b183dbda22012f6a2e91/stamp.jpg"", ""points"": []}",flagged/Detected Instances/e4fae1385b5fe0b62af2/image.webp,45,,,2024-06-25 12:42:48.778457
+flagged/Input Image/4c02c2917f07a0a47195/stamp.jpg,airplane,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/f0ed657840030b178009/stamp.jpg"", ""points"": []}",flagged/Detected Instances/16c30614d826227da8a2/image.webp,0,,,2024-06-25 12:43:02.980277
+flagged/Input Image/965aa552db7559607920/stamp.jpg,horse,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/bab35fa8d8c596d72c83/stamp.jpg"", ""points"": []}",flagged/Detected Instances/eb1276c9005a0e28ff80/image.webp,1,,,2024-06-25 12:43:22.910300
+flagged/Input Image/1f47a5922213f28e3d47/stamp.jpg,green,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/cabeb714ed1e1cddefa1/stamp.jpg"", ""points"": []}",flagged/Detected Instances/fe42b78529616b432d55/image.webp,45,,,2024-06-25 12:43:45.012330
+flagged/Input Image/37cf51da176e45f176c8/lego.jpg,yellow,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/beeb08664504d784a62e/lego.jpg"", ""points"": []}",flagged/Detected Instances/c420149b5762709cd085/image.webp,23,,,2024-06-25 12:44:27.177000
+flagged/Input Image/96219bcbf20113a0aab8/lego.jpg,studs,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/699acba2f203aa14fc16/lego.jpg"", ""points"": []}",flagged/Detected Instances/18d8ce74eab9759ffa86/image.webp,23,,,2024-06-25 12:44:37.787345
+flagged/Input Image/3fa513923f4a4cbc1650/strawberry.jpg,strawberry,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/b26302259971622261f7/strawberry.jpg"", ""points"": []}",flagged/Detected Instances/8d551fe44abbcf336b71/image.webp,16,,,2024-06-25 12:45:02.003915
+flagged/Input Image/d9e2d9688dc9a82fa656/strawberry.jpg,elderberry,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/f86eed288abddd19b0d2/strawberry.jpg"", ""points"": []}",flagged/Detected Instances/cc78b0d8c3d24a3b12e7/image.webp,0,,,2024-06-25 12:45:16.103419
+flagged/Input Image/d74aa8ff2ec0dbba88b7/strawberry.jpg,blueberry,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/89b0adc1664a53a48fea/strawberry.jpg"", ""points"": []}",flagged/Detected Instances/b97b8f4849493ce67de8/image.webp,137,,,2024-06-25 12:45:29.869576
+flagged/Input Image/84eb8736bc8eb48f507f/strawberry.jpg,blueberry,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/0978ec84da8050ca0792/strawberry.jpg"", ""points"": [[330.0, 158.0, 2.0, 354.0, 195.0, 3.0], [289.0, 248.0, 2.0, 307.0, 284.0, 3.0]]}",flagged/Detected Instances/923571e5c8db86d8a8ed/image.webp,132,,,2024-06-25 12:46:16.470770
+flagged/Input Image/104d1b72504a1965f84d/fish.jpg,fish,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/1e96cf74eb0dea109d9f/fish.jpg"", ""points"": []}",flagged/Detected Instances/3df418860ba1a58017cc/image.webp,248,,,2024-06-25 12:46:35.001964
+flagged/Input Image/6ad4c8683b86783f3ec2/stamp.jpg,square,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/1eefba3eed3b8e2a271b/stamp.jpg"", ""points"": [[300.0, 285.0, 2.0, 351.0, 327.0, 3.0], [240.0, 337.0, 2.0, 290.0, 379.0, 3.0]]}",flagged/Detected Instances/eaab5820db88aac49462/image.webp,45,,,2024-06-25 12:47:11.901813
+flagged/Input Image/9eb585667d874ba79393/stamp.jpg,red rectangular stamp,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/88c962fe4751ec3839cd/stamp.jpg"", ""points"": [[300.0, 285.0, 2.0, 351.0, 327.0, 3.0], [240.0, 337.0, 2.0, 290.0, 379.0, 3.0]]}",flagged/Detected Instances/cd25d7af77d301a1d5a4/image.webp,45,,,2024-06-25 12:49:38.403498
+flagged/Input Image/9724ddc2f18fd68bdee9/stamp.jpg,red ,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/13e9c5b34604bb92d7fc/stamp.jpg"", ""points"": [[300.0, 285.0, 2.0, 351.0, 327.0, 3.0], [240.0, 337.0, 2.0, 290.0, 379.0, 3.0]]}",flagged/Detected Instances/57bf0944ab05bd0099ea/image.webp,45,,,2024-06-25 12:49:51.522005
+flagged/Input Image/1238680c118136bb0496/lego.jpg,red lego,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/05be72fc352ad2ba08ba/lego.jpg"", ""points"": []}",flagged/Detected Instances/11c893cbcf45bcf1ab51/image.webp,23,,,2024-06-25 12:50:27.240856
+flagged/Input Image/a305b9fbd30361d8d3bf/lego.jpg,red ,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/a7ebb8ee6c2ad9540939/lego.jpg"", ""points"": []}",flagged/Detected Instances/e5780fbd02076a52113b/image.webp,23,,,2024-06-25 12:50:37.774851
+flagged/Input Image/34d0773dc8f14c58f797/lego.jpg,white,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/23ead2166dba8533b130/lego.jpg"", ""points"": []}",flagged/Detected Instances/91146d3e4f8f27b15856/image.webp,23,,,2024-06-25 12:50:50.401975
+flagged/Input Image/c76b4273734935f64ecf/strawberry.jpg,strawberry,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/ee8ba6f40aa26158d870/strawberry.jpg"", ""points"": []}",flagged/Detected Instances/d1e59cae00d004db8b42/image.webp,16,,,2024-06-25 12:51:05.753104
+flagged/Input Image/a6c4912aa51cc54ed3d7/strawberry.jpg,blueberry,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/566190bf535ad3281f6c/strawberry.jpg"", ""points"": []}",flagged/Detected Instances/0a2ffbe024c77e7da50d/image.webp,137,,,2024-06-25 12:51:31.117635
+flagged/Input Image/ea56a8e9ad4e135ff663/strawberry.jpg,blueberry,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/c92e960c7b3df79ff7b5/strawberry.jpg"", ""points"": [[341.0, 123.0, 2.0, 361.0, 149.0, 3.0]]}",flagged/Detected Instances/428433149285aee052cf/image.webp,134,,,2024-06-25 12:51:42.720447
+flagged/Input Image/6fc4ed98f8495057126d/women.jpg,girl,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/ec38f075c8bc9249bfb6/women.jpg"", ""points"": []}",flagged/Detected Instances/bc7f57f26e088938c86a/image.webp,5,,,2024-06-25 12:52:16.523201
+flagged/Input Image/8d06bc9272737637658c/women.jpg,boy,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/94a9eef129e73dec7ed8/women.jpg"", ""points"": []}",flagged/Detected Instances/c419565161b7941b9368/image.webp,1,,,2024-06-25 12:52:34.380793
+flagged/Input Image/18639d94ba6ee7673454/fish.jpg,fish,"{""image"": ""flagged/CLICK  DRAG to specify object to count with visual exemplars here/4d917d8cff994b9757bd/fish.jpg"", ""points"": []}",flagged/Detected Instances/43bbba43eb4a991b6256/image.webp,248,,,2024-06-25 12:53:25.668942
+flagged/Input Image/acbbf9f607925c34aec9/strawberry.jpg,strawberry,"{""image"": ""flagged/Specify object to count with visual exemplars here/a66486911b04c6d3dd5f/strawberry.jpg"", ""points"": []}",flagged/Detected Instances/71aa71914a32c98dbf5e/image.webp,16,,,2024-06-25 13:21:35.534920
+flagged/Input Image/5a125c995d0990aaf69b/bird-1.JPG,bird,"{""image"": ""flagged/Specify object to count with visual exemplars here/c8b24b1c09de0bbcfa35/bird-2.JPG"", ""points"": []}",flagged/Detected Instances/a24ce3527db178c02415/image.webp,30,,,2024-06-25 13:24:10.446815
+flagged/Input Image/be297ebfd1bb43cd1438/strawberry.jpg,strawberry,"{""image"": ""flagged/Specify object to count with visual exemplars here/d6e9603b839cb704d1e1/strawberry.jpg"", ""points"": []}",flagged/Detected Instances/ab49e40bc9aefd394529/image.webp,16,,,2024-06-25 13:27:08.069932
+flagged/Input Image/667eae2d1cf1f7cb10bf/apple.jpg,apple,"{""image"": ""flagged/Specify object to count with visual exemplars here/d4e8125f9602a8823edd/apple.jpg"", ""points"": []}",flagged/Detected Instances/9568219f942093ef51cf/image.webp,568,,,2024-06-25 13:47:38.683788
+flagged/Input Image/ff98881ffa5a95d08f37/women.jpg,girl,"{""image"": ""flagged/Specify object to count with visual exemplars here/8afee605e88ec6ddaa37/women.jpg"", ""points"": []}",flagged/Detected Instances/9c09433356e51d42d738/image.webp,5,,,2024-06-25 13:47:53.014232
+flagged/Input Image/0c059862756e0017eb5b/deer.jpg,deer,"{""image"": ""flagged/Specify object to count with visual exemplars here/454665f365b4cd0826e9/deer.jpg"", ""points"": []}",flagged/Detected Instances/95bf56b70e2f2e6f3fe6/image.webp,21,,,2024-06-25 13:48:07.565103
+flagged/Input Image/59a8de455bfa9769d8f0/egg.jpg,egg,"{""image"": ""flagged/Specify object to count with visual exemplars here/6626b7927b70e282f908/egg.jpg"", ""points"": []}",flagged/Detected Instances/043ac97ae83b87ef6d7d/image.webp,18,,,2024-06-25 13:48:24.214285
+flagged/Input Image/f916803fd0379cb3b3b6/women.jpg,foot,"{""image"": ""flagged/Specify object to count with visual exemplars here/e743db45e1b6c6dbce71/women.jpg"", ""points"": []}",flagged/Detected Instances/9906067993749438f371/image.webp,0,,,2024-06-25 13:52:46.735106
+flagged/Input Image/b014d0dd065e2cc85351/women.jpg,shoe,"{""image"": ""flagged/Specify object to count with visual exemplars here/15cc3b4b2ee3fa63144b/women.jpg"", ""points"": []}",flagged/Detected Instances/2a54e56bffa5f3c38df3/image.webp,13,,,2024-06-25 13:52:55.287490
+flagged/Input Image/e58990d9da6323e900ec/strawberry.jpg,blueberry,"{""image"": ""flagged/Specify object to count with visual exemplars here/f9d6362f7f07cc8455ba/strawberry.jpg"", ""points"": []}",flagged/Detected Instances/962b404f01817d4b4240/image.webp,137,,,2024-06-25 13:53:24.831821
+flagged/Input Image/36b16cc3616ed8f56ff1/strawberry.jpg,blueberry,"{""image"": ""flagged/Specify object to count with visual exemplars here/c70c72e5fad645bc14aa/strawberry.jpg"", ""points"": [[300.0, 97.0, 2.0, 321.0, 122.0, 3.0]]}",flagged/Detected Instances/c8d02635413c34abe79e/image.webp,134,,,2024-06-25 13:53:33.382644
+flagged/Input Image/abf2935032503027b2a6/strawberry.jpg,,"{""image"": ""flagged/Specify object to count with visual exemplars here/9d3f4b6f03f0dca04bc9/strawberry.jpg"", ""points"": [[300.0, 97.0, 2.0, 321.0, 122.0, 3.0], [165.0, 74.0, 2.0, 204.0, 133.0, 3.0], [104.0, 182.0, 2.0, 151.0, 230.0, 3.0]]}",flagged/Detected Instances/0e52e648edb9fd8db2d5/image.webp,143,,,2024-06-25 13:54:00.465000
+flagged/Input Image/af6c4808a13098046947/strawberry.jpg,,"{""image"": ""flagged/Specify object to count with visual exemplars here/75ed6a32a8a1e6340110/strawberry.jpg"", ""points"": [[99.0, 139.0, 2.0, 146.0, 185.0, 3.0], [161.0, 71.0, 2.0, 202.0, 132.0, 3.0]]}",flagged/Detected Instances/db51c4253db3807889f0/image.webp,16,,,2024-06-25 13:54:22.839814
+flagged/Input Image/c716b9c2502379b4f53d/strawberry.jpg,strawberry,"{""image"": ""flagged/Specify object to count with visual exemplars here/1ab41b769030db4ba746/strawberry.jpg"", ""points"": []}",flagged/Detected Instances/1d060462444efcff7c1b/image.webp,16,,,2024-06-25 14:24:43.543530
+flagged/Input Image/aca09cc83af42dc6a5ed/women.jpg,girl,"{""image"": ""flagged/Specify object to count with visual exemplars here/3399db8186d98f479b4f/women.jpg"", ""points"": []}",flagged/Detected Instances/a878c25daaecd95cd2e3/image.webp,5,,,2024-06-25 14:26:04.205110
+flagged/Input Image/f0e3038613138f5653fe/fish.jpg,fish,"{""image"": ""flagged/Specify object to count with visual exemplars here/5a2fbfebe22860e70887/fish.jpg"", ""points"": []}",flagged/Detected Instances/b00af2ba550cbce038ed/image.webp,248,,,2024-06-25 14:41:36.984087
+flagged/Input Image/f22235f8e59c137765af/deer.jpg,deer,"{""image"": ""flagged/Specify object to count with visual exemplars here/79acfe92d19d88df7924/deer.jpg"", ""points"": []}",flagged/Detected Instances/43e430b66d190a4be7b6/image.webp,21,,,2024-06-25 14:42:57.479610
+flagged/Input Image/43828e990ec1f6c5dc14/fish.jpg,fish,"{""image"": ""flagged/Specify object to count with visual exemplars here/ee76194fb63611ab5b90/fish.jpg"", ""points"": []}",flagged/Detected Instances/04969bb2604f9248742b/image.webp,248,,,2024-06-25 14:45:20.424378
+flagged/Input Image/1e2bb41e969d56102dc2/bird-1.JPG,bird,"{""image"": ""flagged/Specify object to count with visual exemplars here/01989c9431d1e6c77ad1/bird-2.JPG"", ""points"": []}",flagged/Detected Instances/4957f1a733161da1ddcd/image.webp,30,,,2024-06-25 14:48:39.348196
+flagged/Input Image/e1f55639b7c03f2c3bd1/deer.jpg,deer,"{""image"": ""flagged/Specify object to count with visual exemplars here/7207f6ffa7c700e4339b/deer.jpg"", ""points"": []}",flagged/Detected Instances/c64f5cf630c80bfc90b5/image.webp,21,,,2024-06-25 14:48:56.962720
+flagged/Input Image/adef57f83578033397ef/strawberry.jpg,strawberry,"{""image"": ""flagged/Specify object to count with visual exemplars here/a77ac336261b1cb69b31/strawberry.jpg"", ""points"": []}",flagged/Detected Instances/99b7039bc665946889cf/image.webp,16,,,2024-06-25 14:49:23.944300
+flagged/Input Image/fad7be92b6af8fc44818/strawberry.jpg,strawberry,"{""image"": ""flagged/Specify object to count with visual exemplars here/fbd49488b1c780e4d5d1/strawberry.jpg"", ""points"": []}",flagged/Detected Instances/26fa7fa2475e60ea3221/image.webp,16,,,2024-06-25 14:51:48.832356
+flagged/Input Image/ab0e307953f58465f254/strawberry.jpg,strawberry,"{""image"": ""flagged/Specify object to count with visual exemplars here/162fa100594505627551/strawberry.jpg"", ""points"": []}",flagged/Detected Instances/4351f6b9dc586a7a192f/image.webp,16,,,2024-06-25 14:53:19.317939
+flagged/Input Image/64cf922799f8f994f127/strawberry.jpg,strawberry,"{""image"": ""flagged/Specify object to count with visual exemplars here/be38b70786240819da5e/strawberry.jpg"", ""points"": [[183.0, 131.0, 1.0, 0.0, 0.0, 4.0], [272.0, 124.0, 1.0, 0.0, 0.0, 4.0], [272.0, 124.0, 2.0, 325.0, 186.0, 3.0]]}",flagged/Detected Instances/7e5c71b26561217570d0/image.webp,16,,,2024-06-25 14:53:37.751975
+flagged/Input Image/7a7dacbf8c123986c050/strawberry.jpg,strawberry,"{""image"": ""flagged/Specify object to count with visual exemplars here/50f292a91824f56a78ec/strawberry.jpg"", ""points"": [[183.0, 131.0, 1.0, 0.0, 0.0, 4.0], [272.0, 124.0, 1.0, 0.0, 0.0, 4.0], [272.0, 124.0, 2.0, 325.0, 186.0, 3.0], [335.0, 170.0, 1.0, 0.0, 0.0, 4.0], [286.0, 172.0, 1.0, 0.0, 0.0, 4.0], [275.0, 193.0, 1.0, 0.0, 0.0, 4.0], [267.0, 207.0, 2.0, 285.0, 254.0, 3.0], [317.0, 223.0, 2.0, 345.0, 277.0, 3.0]]}",flagged/Detected Instances/cdeab185b16c9564290b/image.webp,150,,,2024-06-25 14:53:51.580488
+flagged/Input Image/4a1ed2bed23e714c2e98/fish.jpg,fish,"{""image"": ""flagged/Specify object to count with visual exemplars here/7b43211f1ed9441e8e60/fish.jpg"", ""points"": [[526.0, 469.0, 2.0, 577.0, 520.0, 3.0]]}",flagged/Detected Instances/b6581780403fac1fe05a/image.webp,247,,,2024-06-25 14:54:23.273247
+flagged/Input Image/c82e8b7773224a87ab1a/strawberry.jpg,blueberry,"{""image"": ""flagged/Specify object to count with visual exemplars here/3f1c6c4b8fa8b0faec98/strawberry.jpg"", ""points"": []}",flagged/Detected Instances/f9ffff38f04c59e61004/image.webp,137,,,2024-06-25 14:56:52.255423
+flagged/Input Image/5e7b302c38bef4e986e9/strawberry.jpg,blueberry,"{""image"": ""flagged/Specify object to count with visual exemplars here/296b7b929afa36c48677/strawberry.jpg"", ""points"": [[329.0, 168.0, 2.0, 353.0, 190.0, 3.0]]}",flagged/Detected Instances/1a9940b04abede2a55a9/image.webp,133,,,2024-06-25 14:57:03.171030
+flagged/Input Image/288e68a5d5996e695512/strawberry.jpg,strawberry,"{""image"": ""flagged/Specify object to count with visual exemplars here/e7aa3e19a9c873c247b5/strawberry.jpg"", ""points"": [[208.0, 92.0, 2.0, 236.0, 139.0, 3.0]]}",flagged/Detected Instances/062ea676672533bd63af/image.webp,16,,,2024-06-25 15:01:55.834836
+flagged/Input Image/65771625dd3209ab4b5e/women.jpg,girl,"{""image"": ""flagged/Specify object to count with visual exemplars here/9cc7376257b49f9a02ef/women.jpg"", ""points"": [[267.0, 320.0, 2.0, 300.0, 362.0, 3.0]]}",flagged/Detected Instances/3e86fdbcad5cb3dbab9a/image.webp,6,,,2024-06-25 15:02:27.623802
+flagged/Input Image/6d2d8805808abba5c681/strawberry.jpg,strawberry,"{""image"": ""flagged/Specify object to count with visual exemplars here/9f49a41bd458e9c37e95/strawberry.jpg"", ""points"": [[305.0, 187.0, 1.0, 0.0, 0.0, 4.0], [166.0, 56.0, 2.0, 203.0, 136.0, 3.0]]}",flagged/Detected Instances/9618d50e52b98e8ecfd1/image.webp,16,,,2024-06-25 15:08:35.121868
+flagged/Input Image/12e08ca8d302970a2008/strawberry.jpg,strawberry,"{""image"": ""flagged/Specify object to count with visual exemplars here/ee9d14118e586de08ead/strawberry.jpg"", ""points"": [[305.0, 187.0, 1.0, 0.0, 0.0, 4.0], [166.0, 56.0, 2.0, 203.0, 136.0, 3.0], [301.0, 96.0, 2.0, 319.0, 118.0, 3.0], [313.0, 165.0, 2.0, 333.0, 184.0, 3.0], [271.0, 282.0, 2.0, 285.0, 309.0, 3.0], [301.0, 242.0, 2.0, 320.0, 262.0, 3.0], [329.0, 162.0, 2.0, 358.0, 185.0, 3.0]]}",flagged/Detected Instances/05fd3125ccda9546e2b6/image.webp,146,,,2024-06-25 15:08:50.604743
+flagged/Input Image/4e1dd60a00267ba7ac80/women.jpg,boy,"{""image"": ""flagged/Specify object to count with visual exemplars here/4a178f28ec8e9901388e/women.jpg"", ""points"": []}",flagged/Detected Instances/b7d4a5530dfdf01f426a/image.webp,1,,,2024-06-25 15:09:42.749784
+flagged/Input Image/813b6b4e4a70fea8e13c/egg.jpg,egg,"{""image"": ""flagged/Specify object to count with visual exemplars here/9fc0bb0e2e5ed2a7c023/egg.jpg"", ""points"": []}",flagged/Detected Instances/6a7dbfbbbf65a567057e/image.webp,18,,,2024-06-25 15:09:55.901192
+flagged/Input Image/c09c36810b55389abddc/lego.jpg,yellow lego stud,"{""image"": ""flagged/Specify object to count with visual exemplars here/696b3fd37ad6c635baba/lego.jpg"", ""points"": []}",flagged/Detected Instances/60c9fcde0bb55251c8f1/image.webp,23,,,2024-06-25 15:10:20.114086
+flagged/Input Image/bd286866d8e9a3e48bad/lego.jpg,red,"{""image"": ""flagged/Specify object to count with visual exemplars here/719beaa78bcd4eb92c8b/lego.jpg"", ""points"": []}",flagged/Detected Instances/87cfe221a685444fa193/image.webp,23,,,2024-06-25 15:10:33.233397
+flagged/Input Image/89c65135bd1fcf0062af/green-pea.jpg,green pea,"{""image"": ""flagged/Specify object to count with visual exemplars here/b2de1a09bc58b6b4d3ae/green-pea.jpg"", ""points"": []}",flagged/Detected Instances/7272d7039ec368ccb1ac/image.webp,213,,,2024-06-25 15:10:42.163670
+flagged/Input Image/a6c33d85c85f75ef8866/stamp.jpg,stamp,"{""image"": ""flagged/Specify object to count with visual exemplars here/679009966f23256eed77/stamp.jpg"", ""points"": []}",flagged/Detected Instances/946f4a626a568a22de21/image.webp,45,,,2024-06-25 15:10:57.228588
+flagged/Input Image/5954d7cfd21da1b1d083/webcam.png,girl,"{""image"": ""flagged/Specify object to count with visual exemplars here/406bdbe8b94e97f309e1/stamp.jpg"", ""points"": []}",flagged/Detected Instances/671208d6944225443700/image.webp,1,,,2024-06-25 15:13:06.226068
+flagged/Input Image/0d653baaf7b60d4fa9a9/webcam.png,,"{""image"": ""flagged/Specify object to count with visual exemplars here/4961820b4b19fa1543ab/webcam.png"", ""points"": [[348.0, 198.0, 2.0, 825.0, 594.0, 3.0]]}",flagged/Detected Instances/11d3ce706b7c55e4b29d/image.webp,6,,,2024-06-25 15:13:46.575188
+flagged/Input Image/2fc47796e17477b72a7b/webcam.png,niki,"{""image"": ""flagged/Specify object to count with visual exemplars here/ed991be65bcc28198008/webcam.png"", ""points"": [[348.0, 198.0, 2.0, 825.0, 594.0, 3.0]]}",flagged/Detected Instances/4b4c120051c76b356ad9/image.webp,11,,,2024-06-25 15:14:11.677104
+flagged/Input Image/c97af0e3fe6b528a4787/webcam.png,girl,"{""image"": ""flagged/Specify object to count with visual exemplars here/b2e5b5ad196b52f66b65/webcam.png"", ""points"": [[348.0, 198.0, 2.0, 825.0, 594.0, 3.0]]}",flagged/Detected Instances/99c1019f28a1ea86e661/image.webp,5,,,2024-06-25 15:14:22.291803
+flagged/Input Image/6d42f7157ab42c91e2d7/webcam.png,girl,"{""image"": ""flagged/Specify object to count with visual exemplars here/7fb61e9fc487a3698878/webcam.png"", ""points"": []}",flagged/Detected Instances/bbe44c1a92d592ae7d96/image.webp,1,,,2024-06-25 15:14:34.641650
+flagged/Input Image/cd2fd26787e15861e3d8/apple.jpg,apple,"{""image"": ""flagged/Specify object to count with visual exemplars here/67cf19d7f6ecdd3495a0/apple.jpg"", ""points"": []}",flagged/Detected Instances/c828265963027a846279/image.webp,568,,,2024-06-25 15:39:38.657610
+flagged/Input Image/676bdd4d886064a651ff/apple.jpg,apple,"{""image"": ""flagged/Specify object to count with visual exemplars here/26e0d8f5fa299f54b2dd/apple.jpg"", ""points"": [[316.0, 184.0, 2.0, 371.0, 220.0, 3.0]]}",flagged/Detected Instances/f281b9015813f6849ef0/image.webp,541,,,2024-06-25 15:40:22.068821
+flagged/Input Image/157f9999f8285255f079/apple.jpg,apple,"{""image"": ""flagged/Specify object to count with visual exemplars here/9f08512a7bc78db40a16/apple.jpg"", ""points"": [[316.0, 184.0, 2.0, 371.0, 220.0, 3.0], [118.0, 245.0, 2.0, 172.0, 276.0, 3.0]]}",flagged/Detected Instances/20546e887359f9ab6311/image.webp,552,,,2024-06-25 15:40:31.337257
+flagged/Input Image/dfefe4e8ad7004a21057/apple.jpg,apple,"{""image"": ""flagged/Specify object to count with visual exemplars here/0e01c3339d835fc181f9/apple.jpg"", ""points"": [[316.0, 184.0, 2.0, 371.0, 220.0, 3.0], [118.0, 245.0, 2.0, 172.0, 276.0, 3.0], [314.0, 230.0, 2.0, 357.0, 265.0, 3.0]]}",flagged/Detected Instances/00f422daee514dc14f95/image.webp,540,,,2024-06-25 15:40:39.970940
+flagged/Input Image/740c54384733c6909520/deer.jpg,deer,"{""image"": ""flagged/Specify object to count with visual exemplars here/20146d57e02b87830aeb/deer.jpg"", ""points"": []}",flagged/Detected Instances/5f4bb2b3b91635f81795/image.webp,21,,,2024-06-25 15:40:53.851874
+flagged/Input Image/5a9d46d5f1fb47ee307b/orchard.jpg,apple,"{""image"": ""flagged/Specify object to count with visual exemplars here/6d352f6e7b5a6c866fdb/deer.jpg"", ""points"": []}",flagged/Detected Instances/6c2b985de60b38896d84/image.webp,543,,,2024-06-25 15:42:01.647977
+flagged/Input Image/de66ddbd6870a9789205/webcam.png,bunny,"{""image"": ""flagged/Specify object to count with visual exemplars here/cd0638cb70d92342ecda/deer.jpg"", ""points"": []}",flagged/Detected Instances/dd795b53a5c38aa3207c/image.webp,0,,,2024-06-25 15:43:54.093401
+flagged/Input Image/4193d822e105ac4695ff/webcam.png,man,"{""image"": ""flagged/Specify object to count with visual exemplars here/13affadc040616f325fb/deer.jpg"", ""points"": []}",flagged/Detected Instances/deb096b5bb87593f8894/image.webp,1,,,2024-06-25 15:44:03.789939
+flagged/Input Image/5dca2e19924d68ff12c5/webcam.png,yoga mat,"{""image"": ""flagged/Specify object to count with visual exemplars here/3c08834b2231ee8173b6/deer.jpg"", ""points"": []}",flagged/Detected Instances/4ccd7a4d047f6cc4f438/image.webp,0,,,2024-06-25 15:44:18.374825
+flagged/Input Image/7dc8361a1e198ce87161/webcam.png,microwave,"{""image"": ""flagged/Specify object to count with visual exemplars here/2ab8fa7854e2e8f488b7/deer.jpg"", ""points"": []}",flagged/Detected Instances/5a1c457670ba55bd7f22/image.webp,1,,,2024-06-25 15:44:29.556338
+flagged/Input Image/e110ca4a4a40edaa6c2f/webcam.png,eyebrow,"{""image"": ""flagged/Specify object to count with visual exemplars here/2b5d4f4150c5af42be5e/deer.jpg"", ""points"": []}",flagged/Detected Instances/e6e0f246c363ee76f378/image.webp,0,,,2024-06-25 15:44:54.438375
+flagged/Input Image/b9323d538671be8fb66c/webcam.png,ear,"{""image"": ""flagged/Specify object to count with visual exemplars here/d218867fab61d8837d1b/deer.jpg"", ""points"": []}",flagged/Detected Instances/8d6a4f375884a2862603/image.webp,0,,,2024-06-25 15:45:04.003679
+Submit,Clear,flagged/Input Image/207cd608b7c1a6c526bd/strawberry.jpg,strawberry,"{""image"": ""flagged/Specify object to count with visual exemplars here/4dfe49b3bb9109a5112c/strawberry.jpg"", ""points"": []}",flagged/Detected Instances/69dbd8ce1b54204aa311/image.webp,16,,,2024-06-25 17:17:28.470393
+Submit,Clear,flagged/Input Image/ccb9fedaf141f3cb6f69/strawberry.jpg,strawberry,"{""image"": ""flagged/Specify object to count with visual exemplars here/9f5e5147ae36d6c07538/strawberry.jpg"", ""points"": []}",flagged/Detected Instances/cd6f4c50bdf9f10604db/image.webp,16,,,2024-06-25 17:48:13.969532

groundingdino/util/__init__.py

@@ -0,0 +1 @@
+# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved