# Built from https://huggingface.co/spaces/hlydecker/MegaDetector_v5 # Built from https://huggingface.co/spaces/sofmi/MegaDetector_DLClive/blob/main/app.py # Built from https://huggingface.co/spaces/Neslihan/megadetector_dlcmodels/blob/main/app.py import gradio as gr import torch import torchvision from dlclive import DLCLive, Processor from PIL import Image, ImageColor, ImageFont, ImageDraw import numpy as np import math # import json import os import yaml # import pdb ######################################### # Input params FONTS = {'amiko': "font/Amiko-Regular.ttf", 'nature': "font/LoveNature.otf", 'painter':"font/PainterDecorator.otf", 'animals': "font/UncialAnimals.ttf", 'zen': "font/ZEN.TTF"} Megadet_Models = {'md_v5a': "megadet_model/md_v5a.0.0.pt", 'md_v5b': "megadet_model/md_v5b.0.0.pt"} ######################################### # Draw keypoints on image def draw_keypoints_on_image(image, keypoints, map_label_id_to_str, flag_show_str_labels, use_normalized_coordinates=True, font_style='amiko', font_size=8, keypt_color="#ff0000", marker_size='2'): """Draws keypoints on an image. Modified from: https://www.programcreek.com/python/?code=fjchange%2Fobject_centric_VAD%2Fobject_centric_VAD-master%2Fobject_detection%2Futils%2Fvisualization_utils.py Args: image: a PIL.Image object. keypoints: a numpy array with shape [num_keypoints, 2]. map_label_id_to_str: dict with keys=label number and values= label string flag_show_str_labels: boolean to select whether or not to show string labels color: color to draw the keypoints with. Default is red. radius: keypoint radius. Default value is 2. use_normalized_coordinates: if True (default), treat keypoint values as relative to the image. Otherwise treat them as absolute. """ # get a drawing context draw = ImageDraw.Draw(image) im_width, im_height = image.size keypoints_x = [k[0] for k in keypoints] keypoints_y = [k[1] for k in keypoints] # adjust keypoints coords if required if use_normalized_coordinates: keypoints_x = tuple([im_width * x for x in keypoints_x]) keypoints_y = tuple([im_height * y for y in keypoints_y]) # draw ellipses around keypoints and add string labels for i, (keypoint_x, keypoint_y) in enumerate(zip(keypoints_x, keypoints_y)): draw.ellipse([(keypoint_x - marker_size, keypoint_y - marker_size), (keypoint_x + marker_size, keypoint_y + marker_size)], outline=keypt_color, fill=keypt_color) # add string labels around keypoints if flag_show_str_labels: font = ImageFont.truetype(FONTS[font_style], font_size) draw.text((keypoint_x + marker_size, keypoint_y + marker_size),#(0.5*im_width, 0.5*im_height), #------- map_label_id_to_str[i], ImageColor.getcolor(keypt_color, "RGB"), # rgb font=font) ############################################ # Predict detections with MegaDetector v5a model def predict_md(im, mega_model_input, size=640): # resize image g = (size / max(im.size)) # multipl factor to make max size of the image equal to input size im = im.resize((int(x * g) for x in im.size), Image.ANTIALIAS) # resize MD_model = torch.hub.load('ultralytics/yolov5', 'custom', Megadet_Models[mega_model_input]) ## detect objects results = MD_model(im) # inference # vars(results).keys()= dict_keys(['imgs', 'pred', 'names', 'files', 'times', 'xyxy', 'xywh', 'xyxyn', 'xywhn', 'n', 't', 's']) results.render() # updates results.imgs with boxes and labels return results ########################################## def crop_animal_detections(yolo_results, likelihood_th): ## Extract animal crops list_labels_as_str = yolo_results.names # ['animal', 'person', 'vehicle'] list_np_animal_crops = [] # for every image for img, det_array in zip(yolo_results.imgs, yolo_results.xyxy): # for every detection for j in range(det_array.shape[0]): # compute coords around bbox rounded to the nearest integer (for pasting later) xmin_rd = int(math.floor(det_array[j,0])) # int() should suffice? ymin_rd = int(math.floor(det_array[j,1])) xmax_rd = int(math.ceil(det_array[j,2])) ymax_rd = int(math.ceil(det_array[j,3])) pred_llk = det_array[j,4] pred_label = det_array[j,5] # keep animal crops above threshold if (pred_label == list_labels_as_str.index('animal')) and \ (pred_llk >= likelihood_th): area = (xmin_rd, ymin_rd, xmax_rd, ymax_rd) crop = Image.fromarray(img).crop(area) crop_np = np.asarray(crop) # add to list list_np_animal_crops.append(crop_np) return list_np_animal_crops ########################################## def predict_dlc(list_np_crops, kpts_likelihood_th, DLCmodel, dlc_proc): # run dlc thru list of crops dlc_live = DLCLive(DLCmodel, processor=dlc_proc) dlc_live.init_inference(list_np_crops[0]) list_kpts_per_crop = [] np_aux = np.empty((1,3)) # can I avoid hardcoding here? for crop in list_np_crops: # scale crop here? keypts_xyp = dlc_live.get_pose(crop) # third column is llk! # set kpts below threhsold to nan keypts_xyp[keypts_xyp[:,-1] < kpts_likelihood_th,:] = np_aux.fill(np.nan) # add kpts of this crop to list list_kpts_per_crop.append(keypts_xyp) return list_kpts_per_crop ##################################################### def predict_pipeline(img_input, mega_model_input, model_input_str, flag_dlc_only, flag_show_str_labels, bbox_likelihood_th, kpts_likelihood_th, font_style, font_size, keypt_color, marker_size, ): ############################################################ ## Get DLC model and labels as strings # TODO: make a dict as for megadetector if model_input_str == 'full_cat': path_to_DLCmodel = "model/DLC_Cat_resnet_50_iteration-0_shuffle-0" pose_cfg_path = os.path.join(path_to_DLCmodel,'pose_cfg.yaml') elif model_input_str == 'full_dog': path_to_DLCmodel = "model/DLC_Dog_resnet_50_iteration-0_shuffle-0" pose_cfg_path = os.path.join(path_to_DLCmodel,'pose_cfg.yaml') elif model_input_str == 'primate_face': path_to_DLCmodel = "model/DLC_FacialLandmarks_resnet_50_iteration-1_shuffle-1" pose_cfg_path = os.path.join(path_to_DLCmodel,'pose_cfg.yaml') elif model_input_str == 'full_human': path_to_DLCmodel = "model/DLC_human_dancing_resnet_101_iteration-0_shuffle-1" pose_cfg_path = os.path.join(path_to_DLCmodel,'pose_cfg.yaml') elif model_input_str == 'full_macaque': path_to_DLCmodel = "model/DLC_monkey_resnet_50_iteration-0_shuffle-1" pose_cfg_path = os.path.join(path_to_DLCmodel,'pose_cfg.yaml') # extract map label ids to strings # pose_cfg_dict['all_joints'] is a list of one-element lists, with open(pose_cfg_path, "r") as stream: pose_cfg_dict = yaml.safe_load(stream) map_label_id_to_str = dict([(k,v) for k,v in zip([el[0] for el in pose_cfg_dict['all_joints']], pose_cfg_dict['all_joints_names'])]) ############################################################ # ### Run Megadetector md_results = predict_md(img_input, mega_model_input, size=640) #Image.fromarray(results.imgs[0]) ################################################################ # Obtain animal crops for bboxes with confidence above th list_crops = crop_animal_detections(md_results, bbox_likelihood_th) ############################################################## # Run DLC dlc_proc = Processor() # if required: ignore MD crops and run DLC on full image [mostly for testing] if flag_dlc_only: # compute kpts on input img list_kpts_per_crop = predict_dlc([np.asarray(img_input)], kpts_likelihood_th, path_to_DLCmodel, dlc_proc) # draw kpts on input img draw_keypoints_on_image(img_input, list_kpts_per_crop[0], # a numpy array with shape [num_keypoints, 2]. map_label_id_to_str, flag_show_str_labels, use_normalized_coordinates=False, font_style=font_style, font_size=font_size, keypt_color=keypt_color, marker_size=marker_size) return img_input else: # Compute kpts for each crop list_kpts_per_crop = predict_dlc(list_crops, kpts_likelihood_th, path_to_DLCmodel, dlc_proc) # Produce final image img_background = Image.fromarray(md_results.imgs[0]) # img_input or Image.fromarray(md_results.imgs[0])? # Image.fromarray(md_results.imgs[0]) --> (640, 479) # img_input.size ---> (259, 194) # pdb.set_trace() # resize image to match megadetector output # g = (640 / max(img_background.size)) # gain # img_background = img_background.resize((int(x * g) for x in img_background.size), Image.ANTIALIAS) # resize for ic, (np_crop, kpts_crop) in enumerate(zip(list_crops, list_kpts_per_crop)): ## Draw keypts on crop img_crop = Image.fromarray(np_crop) draw_keypoints_on_image(img_crop, kpts_crop, # a numpy array with shape [num_keypoints, 2]. map_label_id_to_str, flag_show_str_labels, use_normalized_coordinates=False, # if True, then I should use md_results.xyxyn for list_kpts_crop font_style=font_style, font_size=font_size, keypt_color=keypt_color, marker_size=marker_size) ## Paste crop in original image # https://pillow.readthedocs.io/en/stable/reference/Image.html#PIL.Image.Image.paste img_background.paste(img_crop, box = tuple([int(t) for t in md_results.xyxy[0][ic,:2]])) return img_background ############################################# # %% # User interface: inputs # Input image gr_image_input = gr.inputs.Image(type="pil", label="Input Image") # Models gr_dlc_model_input = gr.inputs.Dropdown(choices=['full_cat','full_dog', 'primate_face', 'full_human', 'full_macaque'], # choices default='full_cat', # default option type='value', # Type of value to be returned by component. "value" returns the string of the choice selected, "index" returns the index of the choice selected. label='Select DeepLabCut model') gr_mega_model_input = gr.inputs.Dropdown(choices=['md_v5a','md_v5b'], default='md_v5a', # default option type='value', # Type of value to be returned by component. "value" returns the string of the choice selected, "index" returns the index of the choice selected. label='Select MegaDetector model') # Other inputs gr_dlc_only_checkbox = gr.inputs.Checkbox(False, label='Run DLClive only, directly on input image?') gr_str_labels_checkbox = gr.inputs.Checkbox(True, label='Show bodypart labels?') gr_slider_conf_bboxes = gr.inputs.Slider(0,1,.05,0.8, label='Set confidence threshold for animal detections') gr_slider_conf_keypoints = gr.inputs.Slider(0,1,.05,0, label='Set confidence threshold for keypoints') # Data viz gr_keypt_color = gr.ColorPicker(label="choose color for keypoint label") gr_labels_font_style = gr.inputs.Dropdown(choices=['amiko', 'nature', 'painter', 'animals', 'zen'], default='amiko', type='value', label='Select keypoint label font') gr_slider_font_size = gr.inputs.Slider(5,30,1,8, label='Set font size') gr_slider_marker_size = gr.inputs.Slider(0.5,5,0.2,2, label='Set marker size') # list of inputs inputs = [gr_image_input, gr_mega_model_input, gr_dlc_model_input, gr_dlc_only_checkbox, gr_str_labels_checkbox, gr_slider_conf_bboxes, gr_slider_conf_keypoints, gr_labels_font_style, gr_slider_font_size, gr_keypt_color, gr_slider_marker_size, ] #################################################### # %% # User interface: outputs gr_image_output = gr.outputs.Image(type="pil", label="Output Image") outputs = [gr_image_output] ############################################## # User interace: description gr_title = "MegaDetector v5 + DeepLabCut-Live!" gr_description = "Contributed by Sofia Minano, Neslihan Wittek, Nirel Kadzo, VicShaoChih Chiang -- DLC AI Residents 2022..\ This App detects and estimate the pose of animals in camera trap images using MegaDetector v5a + DeepLabCut-live. \ We host models from the DeepLabCut ModelZoo Project\, and two MegaDetector Models. Please carefully check their licensing information if you use this project. The App additionally builds upon on work from hlydecker/MegaDetector_v5 \ sofmi/MegaDetector_DLClive \ Neslihan/megadetector_dlcmodels\." # article = "
This app makes predictions using a YOLOv5x6 model that was trained to detect animals, humans, and vehicles in camera trap images; find out more about the project on GitHub. This app was built by Henry Lydecker but really depends on code and models developed by Ecologize and Microsoft AI for Earth. Find out more about the YOLO model from the original creator, Joseph Redmon. YOLOv5 is a family of compound-scaled object detection models trained on the COCO dataset and developed by Ultralytics, and includes simple functionality for Test Time Augmentation (TTA), model ensembling, hyperparameter evolution, and export to ONNX, CoreML and TFLite. Source code | PyTorch Hub
" examples = [['example/monkey_full.jpg', 'full_macaque', False, True, 0.5, 0.3, 'amiko', 5, 'blue', 3, 'md_v5a'], ['example/dog.jpeg', 'full_dog', False, True, 0.5, 0.05, 'amiko', 5, 'yellow', 3, 'md_v5a'], ['example/cat.jpg', 'full_cat', False, True, 0.5, 0.05, 'amiko', 5, 'purple', 3, 'md_v5a']] ################################################ # %% Define and launch gradio interface demo = gr.Interface(predict_pipeline, inputs=inputs, outputs=outputs, title=gr_title, description=gr_description, examples=examples, theme="huggingface", #live=True ) demo.launch(enable_queue=True, share=True)