Create app.py

app.py

@@ -0,0 +1,110 @@
+# --------------------------------------------------------
+# X-Decoder -- Generalized Decoding for Pixel, Image, and Language
+# Copyright (c) 2022 Microsoft
+# Licensed under The MIT License [see LICENSE for details]
+# Written by Jianwei Yang (jianwyan@microsoft.com), Xueyan Zou (xueyan@cs.wisc.edu)
+# --------------------------------------------------------
+
+import os
+os.system("python -m pip install git+https://github.com/MaureenZOU/detectron2-xyz.git")
+
+import gradio as gr
+import torch
+import argparse
+
+from xdecoder.BaseModel import BaseModel
+from xdecoder import build_model
+from utils.distributed import init_distributed
+from utils.arguments import load_opt_from_config_files
+
+from tasks import *
+
+def parse_option():
+    parser = argparse.ArgumentParser('X-Decoder All-in-One Demo', add_help=False)
+    parser.add_argument('--conf_files', default="configs/xdecoder/svlp_focalt_lang.yaml", metavar="FILE", help='path to config file', )
+    args = parser.parse_args()
+
+    return args
+
+'''
+build args
+'''
+args = parse_option()
+opt = load_opt_from_config_files(args.conf_files)
+opt = init_distributed(opt)
+
+# META DATA
+pretrained_pth_last = os.path.join("xdecoder_focalt_last.pt")
+pretrained_pth_novg = os.path.join("xdecoder_focalt_last_novg.pt")
+
+if not os.path.exists(pretrained_pth_last):
+    os.system("wget {}".format("https://projects4jw.blob.core.windows.net/x-decoder/release/xdecoder_focalt_last.pt"))
+
+if not os.path.exists(pretrained_pth_novg):
+    os.system("wget {}".format("https://projects4jw.blob.core.windows.net/x-decoder/release/xdecoder_focalt_last_novg.pt"))
+
+
+'''
+build model
+'''
+model_last = BaseModel(opt, build_model(opt)).from_pretrained(pretrained_pth_last).eval().cuda()
+
+with torch.no_grad():
+    model_last.model.sem_seg_head.predictor.lang_encoder.get_text_embeddings(["background", "background"], is_eval=True)
+
+'''
+inference model
+'''
+
+@torch.no_grad()
+def inference(image, instruction, *args, **kwargs):
+    image = image.convert("RGB")
+    with torch.autocast(device_type='cuda', dtype=torch.float16):
+        return referring_inpainting_gpt3(model_last, image, instruction, *args, **kwargs)
+
+'''
+launch app
+'''
+
+title = "Instructional Image Editing"
+description = """<p style='text-align: center'> <a href='https://x-decoder-vl.github.io/' target='_blank'>Project Page</a> | <a href='https://arxiv.org/pdf/2212.11270.pdf' target='_blank'>Paper</a> | <a href='https://github.com/microsoft/X-Decoder' target='_blank'>Github Repo</a> | <a href='https://youtu.be/wYp6vmyolqE' target='_blank'>Video</a></p>
+                 <p style='text-align: center; color: red;'> NOTE: This demo is mainly for object-centric instructional image editing! For style transfer please refer to the hero demo <a href='https://huggingface.co/spaces/timbrooks/instruct-pix2pix' target='_blank'>Instruct-Pix2Pix</a></p>
+<p>For faster inference without waiting in queue, you may duplicate the space and upgrade to GPU in settings.
+<br/>
+<a href="https://huggingface.co/spaces/xdecoder/Instruct-X-Decoder?duplicate=true">
+<img style="margin-top: 0em; margin-bottom: 0em" src="https://bit.ly/3gLdBN6" alt="Duplicate Space"></a>
+</p>
+"""
+
+help_text = """
+This demo is leveraging X-Decoder's fine-grained understanding for instruction-based image editing. You can use it to:
+1. Remove object, e.g., remove the dog in the image
+2. Replace object, e.g., change the sky with a mountain
+"""
+
+gr.Markdown(help_text)
+
+inputs = [gr.inputs.Image(type='pil'), gr.Textbox(label="instruction")]
+gr.Interface(
+    fn=inference,
+    inputs=inputs,
+    outputs=[
+        gr.outputs.Image(
+        type="pil",
+        label="edit result"),
+    ],
+    examples=[    
+    ["./images/blue_white_bird.jpg", "change the color of bird's feathers from blue to red."],
+    ["./images/house.jpg", "change the house to a modern one."],
+    ["./images/apples.jpg", "change green apple to a red apple"],
+    ["./images/Furniture_Gateway_02.jpg", "make the sofa to one with leather"],
+    ["./images/girl_and_two_boys.png", "remove the boy with blue backbag"],
+    ["./images/dog.png", "remove the chair"],
+    ["./images/horse.png", "change the sky to mountain"],   
+    ["./images/Magritte_TheSonOfMan.jpg", "remove the green apple"]
+    ],
+    title=title,
+    description=description,
+    allow_flagging='never',
+    cache_examples=True,
+).launch()