update changes

app.py

@@ -10,20 +10,23 @@ from transformers import (
     CLIPTextModel,
     CLIPTextModelWithProjection,
 )
-from diffusers import DDPMScheduler, AutoencoderKL
+from diffusers import DDPMScheduler,AutoencoderKL
 from typing import List
+
 import torch
 import os
 from transformers import AutoTokenizer
+
 import numpy as np
+from utils_mask import get_mask_location
 from torchvision import transforms
 import apply_net
 from preprocess.humanparsing.run_parsing import Parsing
 from preprocess.openpose.run_openpose import OpenPose
-from detectron2.data.detection_utils import convert_PIL_to_numpy, _apply_exif_orientation
+from detectron2.data.detection_utils import convert_PIL_to_numpy,_apply_exif_orientation
 from torchvision.transforms.functional import to_pil_image
 
-# Function Definitions and Initialization (Unchanged Sections)
+
 def pil_to_binary_mask(pil_image, threshold=0):
     np_image = np.array(pil_image)
     grayscale_image = Image.fromarray(np_image).convert("L")
@@ -31,9 +34,9 @@ def pil_to_binary_mask(pil_image, threshold=0):
     mask = np.zeros(binary_mask.shape, dtype=np.uint8)
     for i in range(binary_mask.shape[0]):
         for j in range(binary_mask.shape[1]):
-            if binary_mask[i, j]:
-                mask[i, j] = 1
-    mask = (mask * 255).astype(np.uint8)
+            if binary_mask[i,j] == True :
+                mask[i,j] = 1
+    mask = (mask*255).astype(np.uint8)
     output_mask = Image.fromarray(mask)
     return output_mask
 
@@ -41,20 +44,52 @@ def pil_to_binary_mask(pil_image, threshold=0):
 base_path = 'yisol/IDM-VTON'
 example_path = os.path.join(os.path.dirname(__file__), 'example')
 
-# Loading models
-unet = UNet2DConditionModel.from_pretrained(base_path, subfolder="unet", torch_dtype=torch.float16)
+unet = UNet2DConditionModel.from_pretrained(
+    base_path,
+    subfolder="unet",
+    torch_dtype=torch.float16,
+)
 unet.requires_grad_(False)
-
-tokenizer_one = AutoTokenizer.from_pretrained(base_path, subfolder="tokenizer", revision=None, use_fast=False)
-tokenizer_two = AutoTokenizer.from_pretrained(base_path, subfolder="tokenizer_2", revision=None, use_fast=False)
-
+tokenizer_one = AutoTokenizer.from_pretrained(
+    base_path,
+    subfolder="tokenizer",
+    revision=None,
+    use_fast=False,
+)
+tokenizer_two = AutoTokenizer.from_pretrained(
+    base_path,
+    subfolder="tokenizer_2",
+    revision=None,
+    use_fast=False,
+)
 noise_scheduler = DDPMScheduler.from_pretrained(base_path, subfolder="scheduler")
-text_encoder_one = CLIPTextModel.from_pretrained(base_path, subfolder="text_encoder", torch_dtype=torch.float16)
-text_encoder_two = CLIPTextModelWithProjection.from_pretrained(base_path, subfolder="text_encoder_2", torch_dtype=torch.float16)
-image_encoder = CLIPVisionModelWithProjection.from_pretrained(base_path, subfolder="image_encoder", torch_dtype=torch.float16)
-vae = AutoencoderKL.from_pretrained(base_path, subfolder="vae", torch_dtype=torch.float16)
 
-UNet_Encoder = UNet2DConditionModel_ref.from_pretrained(base_path, subfolder="unet_encoder", torch_dtype=torch.float16)
+text_encoder_one = CLIPTextModel.from_pretrained(
+    base_path,
+    subfolder="text_encoder",
+    torch_dtype=torch.float16,
+)
+text_encoder_two = CLIPTextModelWithProjection.from_pretrained(
+    base_path,
+    subfolder="text_encoder_2",
+    torch_dtype=torch.float16,
+)
+image_encoder = CLIPVisionModelWithProjection.from_pretrained(
+    base_path,
+    subfolder="image_encoder",
+    torch_dtype=torch.float16,
+    )
+vae = AutoencoderKL.from_pretrained(base_path,
+                                    subfolder="vae",
+                                    torch_dtype=torch.float16,
+)
+
+# "stabilityai/stable-diffusion-xl-base-1.0",
+UNet_Encoder = UNet2DConditionModel_ref.from_pretrained(
+    base_path,
+    subfolder="unet_encoder",
+    torch_dtype=torch.float16,
+)
 
 parsing_model = Parsing(0)
 openpose_model = OpenPose(0)
@@ -65,35 +100,38 @@ vae.requires_grad_(False)
 unet.requires_grad_(False)
 text_encoder_one.requires_grad_(False)
 text_encoder_two.requires_grad_(False)
-
-tensor_transfrom = transforms.Compose([transforms.ToTensor(), transforms.Normalize([0.5], [0.5])])
+tensor_transfrom = transforms.Compose(
+            [
+                transforms.ToTensor(),
+                transforms.Normalize([0.5], [0.5]),
+            ]
+    )
 
 pipe = TryonPipeline.from_pretrained(
-    base_path,
-    unet=unet,
-    vae=vae,
-    feature_extractor=CLIPImageProcessor(),
-    text_encoder=text_encoder_one,
-    text_encoder_2=text_encoder_two,
-    tokenizer=tokenizer_one,
-    tokenizer_2=tokenizer_two,
-    scheduler=noise_scheduler,
-    image_encoder=image_encoder,
-    torch_dtype=torch.float16,
+        base_path,
+        unet=unet,
+        vae=vae,
+        feature_extractor= CLIPImageProcessor(),
+        text_encoder = text_encoder_one,
+        text_encoder_2 = text_encoder_two,
+        tokenizer = tokenizer_one,
+        tokenizer_2 = tokenizer_two,
+        scheduler = noise_scheduler,
+        image_encoder=image_encoder,
+        torch_dtype=torch.float16,
 )
 pipe.unet_encoder = UNet_Encoder
 
-# Function for try-on functionality
 @spaces.GPU
-def start_tryon(dict, garm_img, garment_des, is_checked_crop, denoise_steps, seed):
+def start_tryon(dict,garm_img,garment_des,is_checked,is_checked_crop,denoise_steps,seed):
     device = "cuda"
     
     openpose_model.preprocessor.body_estimation.model.to(device)
     pipe.to(device)
     pipe.unet_encoder.to(device)
 
-    garm_img = garm_img.convert("RGB").resize((768,1024))
-    human_img_orig = dict["background"].convert("RGB")
+    garm_img= garm_img.convert("RGB").resize((768,1024))
+    human_img_orig = dict["background"].convert("RGB")    
     
     if is_checked_crop:
         width, height = human_img_orig.size
@@ -109,97 +147,163 @@ def start_tryon(dict, garm_img, garment_des, is_checked_crop, denoise_steps, see
     else:
         human_img = human_img_orig.resize((768,1024))
 
+
+    if is_checked:
+        keypoints = openpose_model(human_img.resize((384,512)))
+        model_parse, _ = parsing_model(human_img.resize((384,512)))
+        mask, mask_gray = get_mask_location('hd', "upper_body", model_parse, keypoints)
+        mask = mask.resize((768,1024))
+    else:
+        mask = pil_to_binary_mask(dict['layers'][0].convert("RGB").resize((768, 1024)))
+        # mask = transforms.ToTensor()(mask)
+        # mask = mask.unsqueeze(0)
+    mask_gray = (1-transforms.ToTensor()(mask)) * tensor_transfrom(human_img)
+    mask_gray = to_pil_image((mask_gray+1.0)/2.0)
+
+
+    human_img_arg = _apply_exif_orientation(human_img.resize((384,512)))
+    human_img_arg = convert_PIL_to_numpy(human_img_arg, format="BGR")
+     
+    
+
+    args = apply_net.create_argument_parser().parse_args(('show', './configs/densepose_rcnn_R_50_FPN_s1x.yaml', './ckpt/densepose/model_final_162be9.pkl', 'dp_segm', '-v', '--opts', 'MODEL.DEVICE', 'cuda'))
+    # verbosity = getattr(args, "verbosity", None)
+    pose_img = args.func(args,human_img_arg)    
+    pose_img = pose_img[:,:,::-1]    
+    pose_img = Image.fromarray(pose_img).resize((768,1024))
+    
     with torch.no_grad():
         # Extract the images
         with torch.cuda.amp.autocast():
-            prompt = "model is wearing " + garment_des
-            negative_prompt = "monochrome, lowres, bad anatomy, worst quality, low quality"
-            (
-                prompt_embeds,
-                negative_prompt_embeds,
-                pooled_prompt_embeds,
-                negative_pooled_prompt_embeds,
-            ) = pipe.encode_prompt(
-                prompt,
-                num_images_per_prompt=1,
-                do_classifier_free_guidance=True,
-                negative_prompt=negative_prompt,
-            )
+            with torch.no_grad():
+                prompt = "model is wearing " + garment_des
+                negative_prompt = "monochrome, lowres, bad anatomy, worst quality, low quality"
+                with torch.inference_mode():
+                    (
+                        prompt_embeds,
+                        negative_prompt_embeds,
+                        pooled_prompt_embeds,
+                        negative_pooled_prompt_embeds,
+                    ) = pipe.encode_prompt(
+                        prompt,
+                        num_images_per_prompt=1,
+                        do_classifier_free_guidance=True,
+                        negative_prompt=negative_prompt,
+                    )
+                                    
+                    prompt = "a photo of " + garment_des
+                    negative_prompt = "monochrome, lowres, bad anatomy, worst quality, low quality"
+                    if not isinstance(prompt, List):
+                        prompt = [prompt] * 1
+                    if not isinstance(negative_prompt, List):
+                        negative_prompt = [negative_prompt] * 1
+                    with torch.inference_mode():
+                        (
+                            prompt_embeds_c,
+                            _,
+                            _,
+                            _,
+                        ) = pipe.encode_prompt(
+                            prompt,
+                            num_images_per_prompt=1,
+                            do_classifier_free_guidance=False,
+                            negative_prompt=negative_prompt,
+                        )
 
-            # No mask is used, set mask_image=None
-            pose_img = tensor_transfrom(human_img).unsqueeze(0).to(device, torch.float16)
-            garm_tensor = tensor_transfrom(garm_img).unsqueeze(0).to(device, torch.float16)
-            generator = torch.Generator(device).manual_seed(seed) if seed is not None else None
-            images = pipe(
-                prompt_embeds=prompt_embeds.to(device, torch.float16),
-                negative_prompt_embeds=negative_prompt_embeds.to(device, torch.float16),
-                pooled_prompt_embeds=pooled_prompt_embeds.to(device, torch.float16),
-                negative_pooled_prompt_embeds=negative_pooled_prompt_embeds.to(device, torch.float16),
-                num_inference_steps=denoise_steps,
-                generator=generator,
-                strength=1.0,
-                pose_img=pose_img.to(device, torch.float16),
-                cloth=garm_tensor.to(device, torch.float16),
-                image=human_img,
-                mask_image=None,  # Bypassing the mask
-                height=1024,
-                width=768,
-                ip_adapter_image=garm_img.resize((768, 1024)),
-                guidance_scale=2.0,
-            )[0]
+
+
+                    pose_img =  tensor_transfrom(pose_img).unsqueeze(0).to(device,torch.float16)
+                    garm_tensor =  tensor_transfrom(garm_img).unsqueeze(0).to(device,torch.float16)
+                    generator = torch.Generator(device).manual_seed(seed) if seed is not None else None
+                    images = pipe(
+                        prompt_embeds=prompt_embeds.to(device,torch.float16),
+                        negative_prompt_embeds=negative_prompt_embeds.to(device,torch.float16),
+                        pooled_prompt_embeds=pooled_prompt_embeds.to(device,torch.float16),
+                        negative_pooled_prompt_embeds=negative_pooled_prompt_embeds.to(device,torch.float16),
+                        num_inference_steps=denoise_steps,
+                        generator=generator,
+                        strength = 1.0,
+                        pose_img = pose_img.to(device,torch.float16),
+                        text_embeds_cloth=prompt_embeds_c.to(device,torch.float16),
+                        cloth = garm_tensor.to(device,torch.float16),
+                        mask_image=mask,
+                        image=human_img, 
+                        height=1024,
+                        width=768,
+                        ip_adapter_image = garm_img.resize((768,1024)),
+                        guidance_scale=2.0,
+                    )[0]
 
     if is_checked_crop:
-        out_img = images[0].resize(crop_size)
-        human_img_orig.paste(out_img, (int(left), int(top)))
-        return human_img_orig
+        out_img = images[0].resize(crop_size)        
+        human_img_orig.paste(out_img, (int(left), int(top)))    
+        return human_img_orig, mask_gray
     else:
-        return images[0]
+        return images[0], mask_gray
+    # return images[0], mask_gray
 
-# Loading example garments and human images
-garm_list = os.listdir(os.path.join(example_path, "cloth"))
-garm_list_path = [os.path.join(example_path, "cloth", garm) for garm in garm_list]
+garm_list = os.listdir(os.path.join(example_path,"cloth"))
+garm_list_path = [os.path.join(example_path,"cloth",garm) for garm in garm_list]
 
-human_list = os.listdir(os.path.join(example_path, "human"))
-human_list_path = [os.path.join(example_path, "human", human) for human in human_list]
+human_list = os.listdir(os.path.join(example_path,"human"))
+human_list_path = [os.path.join(example_path,"human",human) for human in human_list]
 
 human_ex_list = []
 for ex_human in human_list_path:
-    ex_dict = {"background": ex_human, "layers": None, "composite": None}
+    ex_dict= {}
+    ex_dict['background'] = ex_human
+    ex_dict['layers'] = None
+    ex_dict['composite'] = None
     human_ex_list.append(ex_dict)
 
-# Building the Gradio interface
+##default human
+
+
 image_blocks = gr.Blocks().queue()
 with image_blocks as demo:
     with gr.Row():
-        # Column 1: Upload person image
         with gr.Column():
-            gr.Markdown("### Step 1: Upload a person image")
-            imgs = gr.ImageEditor(sources="upload", type="pil", label="Human image", interactive=True)
-            is_checked_crop = gr.Checkbox(label="Use auto-crop & resizing", value=False)
+            imgs = gr.ImageEditor(sources='upload', type="pil", label='Human. Mask with pen or use auto-masking', interactive=True)
+            with gr.Row():
+                is_checked = gr.Checkbox(label="Yes", info="Use auto-generated mask (Takes 5 seconds)",value=True)
+            with gr.Row():
+                is_checked_crop = gr.Checkbox(label="Yes", info="Use auto-crop & resizing",value=False)
 
-            gr.Examples(inputs=imgs, examples_per_page=10, examples=human_ex_list, label="Person image examples")
+            example = gr.Examples(
+                inputs=imgs,
+                examples_per_page=10,
+                examples=human_ex_list
+            )
 
-        # Column 2: Upload garment image
         with gr.Column():
-            gr.Markdown("### Step 2: Upload a garment image")
-            garm_img = gr.Image(label="Garment", sources="upload", type="pil")
-            prompt = gr.Textbox(placeholder="Description of garment (e.g., Short Sleeve Round Neck T-shirts)", show_label=False)
-            gr.Examples(inputs=garm_img, examples_per_page=8, examples=garm_list_path, label="Garment examples")
-
-        # Column 3: Display Results
+            garm_img = gr.Image(label="Garment", sources='upload', type="pil")
+            with gr.Row(elem_id="prompt-container"):
+                with gr.Row():
+                    prompt = gr.Textbox(placeholder="Description of garment ex) Short Sleeve Round Neck T-shirts", show_label=False, elem_id="prompt")
+            example = gr.Examples(
+                inputs=garm_img,
+                examples_per_page=8,
+                examples=garm_list_path) 
         with gr.Column():
-            gr.Markdown("### Step 3: Try-On Results")
-            image_out = gr.Image(label="Final Try-On Output", elem_id="output-img", show_share_button=False)
+            # image_out = gr.Image(label="Output", elem_id="output-img", height=400)
+            image_out = gr.Image(label="Output", elem_id="output-img",show_share_button=False)
+
+
 
-    # Advanced settings and run button
-    with gr.Row():
-        try_button = gr.Button(value="Run Try-On")
-        with gr.Accordion("Advanced Settings", open=False):
-            denoise_steps = gr.Number(label="Denoising Steps", minimum=20, maximum=40, value=30, step=1)
-            seed = gr.Number(label="Seed", minimum=-1, maximum=2147483647, step=1, value=42)
 
-    # Button click event to trigger the try-on function
-    try_button.click(fn=start_tryon, inputs=[imgs, garm_img, prompt, is_checked_crop, denoise_steps, seed], outputs=[image_out])
+    with gr.Column():
+        try_button = gr.Button(value="Try-on")
+        with gr.Accordion(label="Advanced Settings", open=False):
+            with gr.Row():
+                denoise_steps = gr.Number(label="Denoising Steps", minimum=20, maximum=40, value=30, step=1)
+                seed = gr.Number(label="Seed", minimum=-1, maximum=2147483647, step=1, value=42)
+
+
+
+    try_button.click(fn=start_tryon, inputs=[imgs, garm_img, prompt, is_checked,is_checked_crop, denoise_steps, seed], outputs=[image_out], api_name='tryon')
+
+            
+
 
-# Launch the Gradio interface
 image_blocks.launch()
+

src/tryon_pipeline.py

@@ -1587,23 +1587,16 @@ class StableDiffusionXLInpaintPipeline(
         )
         init_image = init_image.to(dtype=torch.float32)
 
-        if mask_image is not None:
-            mask = self.mask_processor.preprocess(
-                mask_image, height=height, width=width, resize_mode=resize_mode, crops_coords=crops_coords
-            )
-        else:
-            mask = None  # No mask provided
-
+        mask = self.mask_processor.preprocess(
+            mask_image, height=height, width=width, resize_mode=resize_mode, crops_coords=crops_coords
+        )
         if masked_image_latents is not None:
             masked_image = masked_image_latents
         elif init_image.shape[1] == 4:
             # if images are in latent space, we can't mask it
             masked_image = None
         else:
-            if mask is not None:
-                masked_image = init_image * (mask < 0.5)
-            else:
-                masked_image = None
+            masked_image = init_image * (mask < 0.5)
 
         # 6. Prepare latent variables
         num_channels_latents = self.vae.config.latent_channels