Update app.py

app.py

@@ -1,138 +1,235 @@
+# -*- coding: utf-8 -*-
+import os
+import time
+import tempfile
+from typing import List, Optional, Tuple
+
 import spaces
 import gradio as gr
 from PIL import Image
+
+import torch
+import numpy as np
+from torchvision import transforms
+from torchvision.transforms.functional import to_pil_image
+
+from huggingface_hub import login, snapshot_download
+
 from src.tryon_pipeline import StableDiffusionXLInpaintPipeline as TryonPipeline
 from src.unet_hacked_garmnet import UNet2DConditionModel as UNet2DConditionModel_ref
 from src.unet_hacked_tryon import UNet2DConditionModel
+
 from transformers import (
     CLIPImageProcessor,
     CLIPVisionModelWithProjection,
     CLIPTextModel,
     CLIPTextModelWithProjection,
+    AutoTokenizer,
 )
-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
+from diffusers import DDPMScheduler, AutoencoderKL
+
 import apply_net
+from utils_mask import get_mask_location
 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 torchvision.transforms.functional import to_pil_image
-
+from detectron2.data.detection_utils import convert_PIL_to_numpy, _apply_exif_orientation
+
+
+# =========================
+# Auth (optional)
+# =========================
+DEMO_USER = os.getenv("DEMO_USER", "").strip()
+DEMO_PASS = os.getenv("DEMO_PASS", "").strip()
+APP_AUTH = (DEMO_USER, DEMO_PASS) if (DEMO_USER and DEMO_PASS) else None
+
+# =========================
+# Garments dataset autoload
+# =========================
+GARMENT_DIR = "garments"
+ALLOWED_EXTS = (".png", ".jpg", ".jpeg", ".webp")
+GARMENTS_DATASET = os.getenv("GARMENTS_DATASET", "").strip()  # e.g. "ArmanRV/armanrv-garments"
+HF_TOKEN = os.getenv("HF_TOKEN", "").strip()
+
+def ensure_garments_downloaded() -> None:
+    """
+    Downloads garments from HF Dataset into ./garments to avoid Space repo 1GB limit.
+    """
+    os.makedirs(GARMENT_DIR, exist_ok=True)
+
+    if HF_TOKEN:
+        try:
+            login(token=HF_TOKEN, add_to_git_credential=False)
+            print("HF login: OK")
+        except Exception as e:
+            print("HF login: FAILED:", str(e)[:200])
+
+    if not GARMENTS_DATASET:
+        print("GARMENTS_DATASET not set. Using local ./garments (if any).")
+        return
+
+    try:
+        # Download snapshot to local garments/ (no symlinks for HF container)
+        snapshot_download(
+            repo_id=GARMENTS_DATASET,
+            repo_type="dataset",
+            local_dir=GARMENT_DIR,
+            local_dir_use_symlinks=False,
+            token=HF_TOKEN if HF_TOKEN else None,
+        )
+        print(f"Garments dataset downloaded: {GARMENTS_DATASET} -> {GARMENT_DIR}/")
+    except Exception as e:
+        print("Garments download FAILED:", str(e)[:300])
+
+def list_garments() -> List[str]:
+    try:
+        files = []
+        for f in os.listdir(GARMENT_DIR):
+            if f.lower().endswith(ALLOWED_EXTS) and not f.startswith("."):
+                files.append(f)
+        files.sort()
+        return files
+    except Exception:
+        return []
+
+def garment_path(filename: str) -> str:
+    return os.path.join(GARMENT_DIR, filename)
+
+def load_garment_pil(filename: str) -> Optional[Image.Image]:
+    if not filename:
+        return None
+    path = garment_path(filename)
+    if not os.path.exists(path):
+        return None
+    try:
+        return Image.open(path).convert("RGB")
+    except Exception:
+        return None
+
+def build_gallery_items(files: List[str]):
+    # (image_path, caption) — caption empty for clean UI
+    return [(garment_path(f), "") for f in files]
+
+
+# =========================
+# Small helpers
+# =========================
+def clamp_int(x, lo, hi):
+    try:
+        x = int(x)
+    except Exception:
+        x = lo
+    return max(lo, min(hi, x))
 
 def pil_to_binary_mask(pil_image, threshold=0):
     np_image = np.array(pil_image)
     grayscale_image = Image.fromarray(np_image).convert("L")
     binary_mask = np.array(grayscale_image) > threshold
-    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] == True :
-                mask[i,j] = 1
-    mask = (mask*255).astype(np.uint8)
-    output_mask = Image.fromarray(mask)
-    return output_mask
-
+    mask = (binary_mask.astype(np.uint8) * 255)
+    return Image.fromarray(mask)
+
+# global simple rate limit (helps avoid spam during internal demo)
+_last_call_ts = 0.0
+def allow_call(min_interval_sec: float = 2.5) -> Tuple[bool, str]:
+    global _last_call_ts
+    now = time.time()
+    if now - _last_call_ts < min_interval_sec:
+        wait = min_interval_sec - (now - _last_call_ts)
+        return False, f"⏳ Подождите {wait:.1f} сек."
+    _last_call_ts = now
+    return True, ""
+
+
+# =========================
+# Model init (local IDM-VTON)
+# =========================
+base_path = "yisol/IDM-VTON"
+
+# device policy
+DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
+DTYPE = torch.float16 if DEVICE == "cuda" else torch.float32
+
+print("DEVICE:", DEVICE, "DTYPE:", DTYPE)
+
+# Load components
+unet = UNet2DConditionModel.from_pretrained(base_path, subfolder="unet", torch_dtype=DTYPE)
+unet.requires_grad_(False)
 
-base_path = 'yisol/IDM-VTON'
-example_path = os.path.join(os.path.dirname(__file__), 'example')
+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)
 
-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,
-)
 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,
-)
+text_encoder_one = CLIPTextModel.from_pretrained(base_path, subfolder="text_encoder", torch_dtype=DTYPE)
+text_encoder_two = CLIPTextModelWithProjection.from_pretrained(base_path, subfolder="text_encoder_2", torch_dtype=DTYPE)
 
-# "stabilityai/stable-diffusion-xl-base-1.0",
-UNet_Encoder = UNet2DConditionModel_ref.from_pretrained(
-    base_path,
-    subfolder="unet_encoder",
-    torch_dtype=torch.float16,
-)
+image_encoder = CLIPVisionModelWithProjection.from_pretrained(base_path, subfolder="image_encoder", torch_dtype=DTYPE)
+
+vae = AutoencoderKL.from_pretrained(base_path, subfolder="vae", torch_dtype=DTYPE)
 
+UNet_Encoder = UNet2DConditionModel_ref.from_pretrained(base_path, subfolder="unet_encoder", torch_dtype=DTYPE)
+UNet_Encoder.requires_grad_(False)
+
+# Parsing/OpenPose init
+# These are heavy; GPU intended. On CPU it may be very slow.
 parsing_model = Parsing(0)
 openpose_model = OpenPose(0)
 
-UNet_Encoder.requires_grad_(False)
-image_encoder.requires_grad_(False)
-vae.requires_grad_(False)
-unet.requires_grad_(False)
-text_encoder_one.requires_grad_(False)
-text_encoder_two.requires_grad_(False)
+# Freeze
+for m in [UNet_Encoder, image_encoder, vae, unet, text_encoder_one, text_encoder_two]:
+    m.requires_grad_(False)
+
 tensor_transfrom = transforms.Compose(
-            [
-                transforms.ToTensor(),
-                transforms.Normalize([0.5], [0.5]),
-            ]
-    )
+    [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=DTYPE,
 )
 pipe.unet_encoder = UNet_Encoder
 
-@spaces.GPU
-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)
+
+# =========================
+# Inference
+# =========================
+@spaces.GPU  # ok on dedicated GPU too
+def start_tryon(
+    human_pil: Image.Image,
+    garm_img: Image.Image,
+    auto_mask: bool = True,
+    crop_center: bool = True,
+    denoise_steps: int = 25,
+    seed: int = 42,
+):
+    """
+    Simplified local try-on.
+    Returns: (output_image, masked_preview)
+    """
+    device = "cuda" if torch.cuda.is_available() else "cpu"
+    dtype = torch.float16 if device == "cuda" else torch.float32
+
+    # move heavy models
+    if 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")    
-    
-    if is_checked_crop:
+    # resize inputs to expected
+    garm_img = garm_img.convert("RGB").resize((768, 1024))
+    human_img_orig = human_pil.convert("RGB")
+
+    # optional center crop
+    if crop_center:
         width, height = human_img_orig.size
         target_width = int(min(width, height * (3 / 4)))
         target_height = int(min(height, width * (4 / 3)))
@@ -142,172 +239,224 @@ def start_tryon(dict,garm_img,garment_des,is_checked,is_checked_crop,denoise_ste
         bottom = (height + target_height) / 2
         cropped_img = human_img_orig.crop((left, top, right, bottom))
         crop_size = cropped_img.size
-        human_img = cropped_img.resize((768,1024))
+        human_img = cropped_img.resize((768, 1024))
     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))
+        human_img = human_img_orig.resize((768, 1024))
+
+    # mask
+    if auto_mask:
+        keypoints = openpose_model(human_img.resize((384, 512)))
+        model_parse, _ = parsing_model(human_img.resize((384, 512)))
+        mask, _ = 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)
+        # if someday you add manual mask, you can pass it here
+        mask = Image.new("L", (768, 1024), 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)))
+    # densepose
+    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():
-            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,
-                        )
-
-
-
-                    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, mask_gray
-    else:
-        return images[0], mask_gray
-    # return images[0], mask_gray
 
-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]
+    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" if device == "cuda" else "cpu",
+    ))
+    pose_img = args.func(args, human_img_arg)
+    pose_img = pose_img[:, :, ::-1]
+    pose_img = Image.fromarray(pose_img).resize((768, 1024))
+
+    # prompts (fixed, like your API demo)
+    garment_des = "a garment"
+    prompt_main = "model is wearing " + garment_des
+    prompt_cloth = "a photo of " + garment_des
+    negative_prompt = "monochrome, lowres, bad anatomy, worst quality, low quality"
+
+    denoise_steps = clamp_int(denoise_steps, 20, 40)
+    seed = clamp_int(seed, 0, 999999)
+
+    # inference
+    with torch.no_grad():
+        if device == "cuda":
+            autocast_ctx = torch.cuda.amp.autocast()
+        else:
+            # no autocast on cpu
+            class _NoCtx:
+                def __enter__(self): return None
+                def __exit__(self, *args): return False
+            autocast_ctx = _NoCtx()
+
+        with autocast_ctx:
+            # encode prompts
+            (
+                prompt_embeds,
+                negative_prompt_embeds,
+                pooled_prompt_embeds,
+                negative_pooled_prompt_embeds,
+            ) = pipe.encode_prompt(
+                prompt_main,
+                num_images_per_prompt=1,
+                do_classifier_free_guidance=True,
+                negative_prompt=negative_prompt,
+            )
 
-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]
+            (
+                prompt_embeds_c,
+                _,
+                _,
+                _,
+            ) = pipe.encode_prompt(
+                [prompt_cloth],
+                num_images_per_prompt=1,
+                do_classifier_free_guidance=False,
+                negative_prompt=[negative_prompt],
+            )
 
-human_ex_list = []
-for ex_human in human_list_path:
-    ex_dict= {}
-    ex_dict['background'] = ex_human
-    ex_dict['layers'] = None
-    ex_dict['composite'] = None
-    human_ex_list.append(ex_dict)
+            pose_t = tensor_transfrom(pose_img).unsqueeze(0).to(device=device, dtype=dtype)
+            garm_t = tensor_transfrom(garm_img).unsqueeze(0).to(device=device, dtype=dtype)
+
+            generator = torch.Generator(device).manual_seed(seed)
+
+            images = pipe(
+                prompt_embeds=prompt_embeds.to(device=device, dtype=dtype),
+                negative_prompt_embeds=negative_prompt_embeds.to(device=device, dtype=dtype),
+                pooled_prompt_embeds=pooled_prompt_embeds.to(device=device, dtype=dtype),
+                negative_pooled_prompt_embeds=negative_pooled_prompt_embeds.to(device=device, dtype=dtype),
+                num_inference_steps=denoise_steps,
+                generator=generator,
+                strength=1.0,
+                pose_img=pose_t,
+                text_embeds_cloth=prompt_embeds_c.to(device=device, dtype=dtype),
+                cloth=garm_t,
+                mask_image=mask,
+                image=human_img,
+                height=1024,
+                width=768,
+                ip_adapter_image=garm_img.resize((768, 1024)),
+                guidance_scale=2.0,
+            )[0]
+
+    out_img = images[0]
+    if crop_center:
+        out_img_rs = out_img.resize(crop_size)
+        human_img_orig.paste(out_img_rs, (int(left), int(top)))
+        return human_img_orig, mask_gray
+    return out_img, mask_gray
+
+
+# =========================
+# UI (API-like)
+# =========================
+CUSTOM_CSS = """
+footer {display:none !important;}
+#api-info {display:none !important;}
+div[class*="footer"] {display:none !important;}
+button[aria-label="Settings"] {display:none !important;}
+"""
+
+def refresh_catalog():
+    ensure_garments_downloaded()
+    files = list_garments()
+    items = build_gallery_items(files)
+    status = "✅ Каталог обновлён" if files else "⚠️ Каталог пуст (dataset не скачался или нет файлов)"
+    return items, files, None, status
+
+def on_gallery_select(files_list: List[str], evt: gr.SelectData):
+    if not files_list:
+        return None, "⚠️ Каталог пуст"
+    idx = int(evt.index) if evt.index is not None else 0
+    idx = max(0, min(idx, len(files_list) - 1))
+    return files_list[idx], f"👕 Выбрано: {files_list[idx]}"
+
+def tryon_ui(person_pil, selected_filename):
+    ok, msg = allow_call(2.5)
+    if not ok:
+        return None, None, msg
+
+    if person_pil is None:
+        return None, None, "❌ Загрузите фото человека"
+    if not selected_filename:
+        return None, None, "❌ Выберите одежду из каталога"
+
+    garm = load_garment_pil(selected_filename)
+    if garm is None:
+        return None, None, "❌ Не удалось загрузить выбранную одежду"
+
+    out, masked = start_tryon(
+        human_pil=person_pil,
+        garm_img=garm,
+        auto_mask=True,
+        crop_center=True,
+        denoise_steps=25,
+        seed=42,
+    )
+    return out, masked, "✅ Готово"
 
-##default human
 
+# ensure garments present at startup (best effort)
+ensure_garments_downloaded()
+_initial_files = list_garments()
+_initial_items = build_gallery_items(_initial_files)
 
-image_blocks = gr.Blocks().queue()
-with image_blocks as demo:
-    gr.Markdown("## IDM-VTON 👕👔👚")
-    gr.Markdown("Virtual Try-on with your image and garment image. Check out the [source codes](https://github.com/yisol/IDM-VTON) and the [model](https://huggingface.co/yisol/IDM-VTON)")
-    with gr.Row():
-        with gr.Column():
-            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)
+with gr.Blocks(title="Virtual Try-On Rendez-vous", css=CUSTOM_CSS) as demo:
+    gr.Markdown("# Virtual Try-On Rendez-vous")
 
-            example = gr.Examples(
-                inputs=imgs,
-                examples_per_page=10,
-                examples=human_ex_list
-            )
+    garment_files_state = gr.State(_initial_files)
+    selected_garment_state = gr.State(None)
 
+    with gr.Row():
         with gr.Column():
-            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():
-            # image_out = gr.Image(label="Output", elem_id="output-img", height=400)
-            masked_img = gr.Image(label="Masked image output", elem_id="masked-img",show_share_button=False)
-        with gr.Column():
-            # 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)
+            person = gr.Image(label="Фото человека", type="pil", height=420)
 
-
-
-
-    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)
-
+                refresh_btn = gr.Button("🔄 Обновить каталог одежды", variant="secondary")
+                selected_label = gr.Markdown("👕 Выберите одежду ниже")
+
+            garment_gallery = gr.Gallery(
+                label="Одежда для примерки",
+                value=_initial_items,
+                columns=4,
+                height=340,
+                allow_preview=True,
+            )
 
+            run = gr.Button("Примерить", variant="primary")
+            status = gr.Textbox(value="Ожидание...", interactive=False)
 
-    try_button.click(fn=start_tryon, inputs=[imgs, garm_img, prompt, is_checked,is_checked_crop, denoise_steps, seed], outputs=[image_out,masked_img], api_name='tryon')
+        with gr.Column():
+            out = gr.Image(label="Результат", type="pil", height=520)
+            masked = gr.Image(label="Маска/предпросмотр (служебное)", type="pil", height=320)
 
-            
+    garment_gallery.select(
+        fn=on_gallery_select,
+        inputs=[garment_files_state],
+        outputs=[selected_garment_state, selected_label],
+    )
 
+    refresh_btn.click(
+        fn=refresh_catalog,
+        inputs=[],
+        outputs=[garment_gallery, garment_files_state, selected_garment_state, status],
+    )
 
-image_blocks.launch()
+    run.click(
+        fn=tryon_ui,
+        inputs=[person, selected_garment_state],
+        outputs=[out, masked, status],
+    )
 
+if __name__ == "__main__":
+    demo.launch(
+        server_name="0.0.0.0",
+        server_port=7860,
+        share=False,
+        auth=APP_AUTH,
+    )