app.py

# -*- coding: utf-8 -*-
import os
import time
import json
from datetime import datetime
from typing import List, Optional, Tuple

import spaces
import gradio as gr
from PIL import Image

# =========================
# FIX: gradio 4.24 / gradio_client crashes on boolean JSON Schemas in /api_info
# =========================
def _patch_gradio_client_bool_schema():
    try:
        import gradio_client.utils as gcu
        patched_any = False

        if hasattr(gcu, "get_type"):
            _orig_get_type = gcu.get_type

            def _get_type_patched(schema):
                if isinstance(schema, bool):
                    return "any"
                return _orig_get_type(schema)

            gcu.get_type = _get_type_patched
            patched_any = True

        if hasattr(gcu, "get_desc"):
            _orig_get_desc = gcu.get_desc

            def _get_desc_patched(schema):
                if isinstance(schema, bool):
                    return ""
                return _orig_get_desc(schema)

            gcu.get_desc = _get_desc_patched
            patched_any = True

        if hasattr(gcu, "_json_schema_to_python_type"):
            _orig_json2py = gcu._json_schema_to_python_type

            def _json_schema_to_python_type_patched(schema, defs=None):
                if isinstance(schema, bool):
                    return "any"
                return _orig_json2py(schema, defs)

            gcu._json_schema_to_python_type = _json_schema_to_python_type_patched
            patched_any = True

        if patched_any:
            print("Patched gradio_client.utils for boolean JSON Schemas (/api_info)", flush=True)
        else:
            print("gradio_client patch: nothing to patch (unexpected utils layout)", flush=True)

    except Exception as e:
        print("gradio_client patch failed:", repr(e), flush=True)


_patch_gradio_client_bool_schema()

import torch
from torchvision import transforms
from huggingface_hub import login, snapshot_download, HfApi, hf_hub_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

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


# =========================
# Garments catalog (only)
# =========================
GARMENT_DIR = "garments"
ALLOWED_EXTS = (".png", ".jpg", ".jpeg", ".webp")
GARMENTS_DATASET = os.getenv("GARMENTS_DATASET", "").strip()
HF_TOKEN = os.getenv("HF_TOKEN", "").strip()


def ensure_garments_available() -> None:
    """
    Если GARMENTS_DATASET не задан — используем локальную папку ./garments.
    Если задан — скачиваем датасет HF в ./garments.
    """
    os.makedirs(GARMENT_DIR, exist_ok=True)

    if not GARMENTS_DATASET:
        print("GARMENTS_DATASET not set. Using local ./garments (if any).", flush=True)
        return

    if HF_TOKEN:
        try:
            login(token=HF_TOKEN, add_to_git_credential=False)
            print("HF login: OK", flush=True)
        except Exception as e:
            print("HF login: FAILED:", str(e)[:200], flush=True)

    try:
        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}/", flush=True)
    except Exception as e:
        print("Garments download FAILED:", str(e)[:300], flush=True)


def _gender_subdir(gender: str) -> str:
    return "male" if gender == "Мужская" else "female"


def list_garments(gender: Optional[str] = None) -> List[str]:
    files: List[str] = []
    if not os.path.isdir(GARMENT_DIR):
        return files

    if gender:
        base = os.path.join(GARMENT_DIR, _gender_subdir(gender))
        if os.path.isdir(base):
            for root, _, fnames in os.walk(base):
                for f in fnames:
                    if f.lower().endswith(ALLOWED_EXTS) and not f.startswith("."):
                        rel = os.path.relpath(os.path.join(root, f), GARMENT_DIR)
                        files.append(rel)
            files.sort()
            return files

    for root, _, fnames in os.walk(GARMENT_DIR):
        for f in fnames:
            if f.lower().endswith(ALLOWED_EXTS) and not f.startswith("."):
                rel = os.path.relpath(os.path.join(root, f), GARMENT_DIR)
                files.append(rel)
    files.sort()
    return files


def garment_path(relpath: str) -> str:
    return os.path.join(GARMENT_DIR, relpath)


def load_garment_pil(relpath: str) -> Optional[Image.Image]:
    if not relpath:
        return None
    path = garment_path(relpath)
    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]):
    return [(garment_path(f), "") for f in files]


# =========================
# Rate limit
# =========================
_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 (BASELINE)
# =========================
base_path = "yisol/IDM-VTON"

DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
DTYPE = torch.float16 if DEVICE == "cuda" else torch.float32
print("DEVICE:", DEVICE, "DTYPE:", DTYPE, flush=True)

tensor_transfrom = transforms.Compose([
    transforms.ToTensor(),
    transforms.Normalize([0.5], [0.5]),
])

unet = UNet2DConditionModel.from_pretrained(base_path, subfolder="unet", torch_dtype=DTYPE)
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=DTYPE)
text_encoder_two = CLIPTextModelWithProjection.from_pretrained(base_path, subfolder="text_encoder_2", torch_dtype=DTYPE)

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_model = Parsing(0)
openpose_model = OpenPose(0)

for m in [UNet_Encoder, image_encoder, vae, unet, text_encoder_one, text_encoder_two]:
    m.requires_grad_(False)

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=DTYPE,
)
pipe.unet_encoder = UNet_Encoder


# =========================
# Inference (BASELINE params)
# =========================
@spaces.GPU
def start_tryon(
    human_pil: Image.Image,
    garm_img: Image.Image,
) -> Image.Image:
    device = "cuda" if torch.cuda.is_available() else "cpu"
    dtype = torch.float16 if device == "cuda" else torch.float32

    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 = human_pil.convert("RGB")

    width, height = human_img_orig.size
    target_width = int(min(width, height * (3 / 4)))
    target_height = int(min(height, width * (4 / 3)))
    left = (width - target_width) / 2
    top = (height - target_height) / 2
    right = (width + target_width) / 2
    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))

    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))

    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" 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))

    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 = 30
    guidance_scale = 2.0
    strength = 1.0
    seed = 42

    with torch.no_grad():
        if device == "cuda":
            autocast_ctx = torch.cuda.amp.autocast()
        else:
            class _NoCtx:
                def __enter__(self):
                    return None

                def __exit__(self, *args):
                    return False

            autocast_ctx = _NoCtx()

        with autocast_ctx:
            (
                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,
            )

            (prompt_embeds_c, _, _, _) = pipe.encode_prompt(
                [prompt_cloth],
                num_images_per_prompt=1,
                do_classifier_free_guidance=False,
                negative_prompt=[negative_prompt],
            )

            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=strength,
                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=guidance_scale,
            )[0]

    out_img = images[0]
    out_img_rs = out_img.resize(crop_size)
    human_img_orig.paste(out_img_rs, (int(left), int(top)))
    return human_img_orig


# =========================
# UI / CSS
# =========================
CUSTOM_CSS = """
footer {display:none !important;}
#api-info {display:none !important;}
div[class*="footer"] {display:none !important;}
button[aria-label="Settings"] {display:none !important;}

.feedback-box {
    border: 1px solid #e5e7eb;
    border-radius: 14px;
    padding: 12px 14px;
    background: #fafafa;
    margin-top: 10px;
}

.feedback-ok {
    border: 1px solid #b7ebc6;
    background: #f0fff4;
    color: #166534;
    border-radius: 12px;
    padding: 10px 12px;
    font-size: 14px;
    margin-top: 8px;
}

.feedback-idle {
    border: 1px dashed #d1d5db;
    background: #ffffff;
    color: #6b7280;
    border-radius: 12px;
    padding: 10px 12px;
    font-size: 14px;
    margin-top: 8px;
}
"""


# =========================
# UX example image
# =========================
UX_EXAMPLE_IMG_PATH = "assets/photo_2026-02-26_14-56-24.jpg"


def _load_ux_example_pil() -> Optional[Image.Image]:
    try:
        if UX_EXAMPLE_IMG_PATH and os.path.exists(UX_EXAMPLE_IMG_PATH):
            return Image.open(UX_EXAMPLE_IMG_PATH).convert("RGB")
    except Exception as e:
        print("UX example image load failed:", repr(e), flush=True)
    return None


_UX_EXAMPLE_PIL = _load_ux_example_pil()


def refresh_catalog(gender: str):
    ensure_garments_available()
    files = list_garments(gender=gender)
    items = build_gallery_items(files)
    status = f"✅ Каталог: {gender} ({len(files)})" if files else f"⚠️ Каталог пуст: {gender}"
    return items, files, None, status, "👕 Выберите одежду ниже"


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], "👕 Одежда выбрана"


# =========================
# Feedback storage
# =========================
FEEDBACK_DIR = "./feedback"
FEEDBACK_PATH = os.path.join(FEEDBACK_DIR, "feedback.jsonl")

FEEDBACK_REPO_ID = os.getenv("FEEDBACK_REPO_ID", "").strip()
FEEDBACK_REPO_TYPE = "dataset"
FEEDBACK_REPO_FILEPATH = "feedback/feedback.jsonl"


def _read_local_feedback_text() -> str:
    if not os.path.exists(FEEDBACK_PATH):
        return ""
    with open(FEEDBACK_PATH, "r", encoding="utf-8") as f:
        return f.read()


def _write_local_feedback_text(text: str) -> None:
    os.makedirs(FEEDBACK_DIR, exist_ok=True)
    with open(FEEDBACK_PATH, "w", encoding="utf-8") as f:
        f.write(text)


def _download_repo_feedback(api: HfApi, token: str) -> str:
    if not FEEDBACK_REPO_ID:
        return ""
    try:
        local_path = hf_hub_download(
            repo_id=FEEDBACK_REPO_ID,
            repo_type=FEEDBACK_REPO_TYPE,
            filename=FEEDBACK_REPO_FILEPATH,
            token=token,
        )
        with open(local_path, "r", encoding="utf-8") as f:
            return f.read()
    except Exception:
        return ""


def _upload_repo_feedback(api: HfApi, token: str, text: str) -> None:
    if not FEEDBACK_REPO_ID:
        raise RuntimeError("FEEDBACK_REPO_ID not set")

    os.makedirs(FEEDBACK_DIR, exist_ok=True)
    tmp_path = os.path.join(FEEDBACK_DIR, "_feedback_upload.jsonl")

    with open(tmp_path, "w", encoding="utf-8") as f:
        f.write(text)

    api.upload_file(
        path_or_fileobj=tmp_path,
        path_in_repo=FEEDBACK_REPO_FILEPATH,
        repo_id=FEEDBACK_REPO_ID,
        repo_type=FEEDBACK_REPO_TYPE,
        token=token,
        commit_message="Add try-on feedback",
    )


def _append_feedback_record(record: dict) -> None:
    line = json.dumps(record, ensure_ascii=False) + "\n"

    try:
        local_text = _read_local_feedback_text()
        local_text += line
        _write_local_feedback_text(local_text)
    except Exception as e:
        print("Feedback local write failed:", repr(e), flush=True)

    try:
        token = (os.getenv("HF_TOKEN", "").strip() or os.getenv("HUGGINGFACEHUB_API_TOKEN", "").strip())
        if not token:
            print("Feedback repo sync skipped: HF_TOKEN not set", flush=True)
            return

        if not FEEDBACK_REPO_ID:
            print("Feedback repo sync skipped: FEEDBACK_REPO_ID not set", flush=True)
            return

        api = HfApi()
        repo_text = _download_repo_feedback(api, token)
        repo_text += line
        _upload_repo_feedback(api, token, repo_text)
    except Exception as e:
        print("Feedback repo sync failed:", repr(e), flush=True)


def save_rating_feedback(is_like: bool, garment_name: str) -> None:
    record = {
        "timestamp": datetime.utcnow().isoformat(),
        "event": "rating",
        "like": bool(is_like),
        "garment": garment_name or "",
    }
    _append_feedback_record(record)


def save_comment_feedback(garment_name: str, comment: str) -> None:
    clean_comment = (comment or "").strip()
    if len(clean_comment) > 1500:
        clean_comment = clean_comment[:1500]

    record = {
        "timestamp": datetime.utcnow().isoformat(),
        "event": "comment",
        "garment": garment_name or "",
        "comment": clean_comment,
    }
    _append_feedback_record(record)


# =========================
# Feedback UI helpers
# =========================
def _rating_notice_idle_html():
    return """
    <div class="feedback-idle">
        Оцените результат: нажмите лайк или дизлайк.
    </div>
    """


def _comment_notice_idle_html():
    return """
    <div class="feedback-idle">
        При желании напишите комментарий и нажмите кнопку отправки.
    </div>
    """


def _rating_notice_ok_html(action_text: str):
    return f"""
    <div class="feedback-ok">
        ✅ Ваша оценка <b style="color:#166534;">{action_text}</b> сохранена.
    </div>
    """


def _comment_notice_ok_html():
    return """
    <div class="feedback-ok">
        ✅ Комментарий отправлен. Спасибо за обратную связь.
    </div>
    """


# =========================
# Feedback actions
# =========================
def submit_like_feedback(garment_name: str):
    if not garment_name:
        return "⚠️ Сначала выполните примерку и выберите одежду", _rating_notice_idle_html()

    try:
        save_rating_feedback(True, garment_name)
        return "✅ Оценка сохранена: «Нравится»", _rating_notice_ok_html("«Нравится»")
    except Exception as e:
        return (
            f"❌ Ошибка сохранения оценки: {type(e).__name__}: {str(e)[:200]}",
            _rating_notice_idle_html(),
        )


def submit_dislike_feedback(garment_name: str):
    if not garment_name:
        return "⚠️ Сначала выполните примерку и выберите одежду", _rating_notice_idle_html()

    try:
        save_rating_feedback(False, garment_name)
        return "✅ Оценка сохранена: «Не нравится»", _rating_notice_ok_html("«Не нравится»")
    except Exception as e:
        return (
            f"❌ Ошибка сохранения оценки: {type(e).__name__}: {str(e)[:200]}",
            _rating_notice_idle_html(),
        )


def submit_comment_feedback(garment_name: str, comment: str):
    clean_comment = (comment or "").strip()

    if not garment_name:
        return (
            "⚠️ Сначала выполните примерку и выберите одежду",
            _comment_notice_idle_html(),
            gr.update(value=comment),
        )

    if not clean_comment:
        return (
            "⚠️ Напишите комментарий перед отправкой",
            _comment_notice_idle_html(),
            gr.update(value=comment),
        )

    try:
        save_comment_feedback(garment_name, clean_comment)
        return (
            "✅ Комментарий отправлен",
            _comment_notice_ok_html(),
            gr.update(value=""),
        )
    except Exception as e:
        return (
            f"❌ Ошибка отправки комментария: {type(e).__name__}: {str(e)[:200]}",
            _comment_notice_idle_html(),
            gr.update(value=comment),
        )


# =========================
# Try-on UI
# =========================
def tryon_ui(person_pil, selected_filename):
    for msg in [
    "🧵 Анализируем посадку ткани…",
    "📏 Подбираем пропорции одежды…",
    "🧍 Определяем позу и положение тела…",
    "📐 Выравниваем геометрию одежды…",
    "🎨 Сохраняем цвет и фактуру ткани…",
    "🪡 Прорисовываем швы и детали…",
    "🧶 Адаптируем складки ткани…",
    "🧥 Корректируем посадку на фигуре…",
    "✨ Улучшаем освещение и тени…",
    "🔬 Уточняем текстуру материала…"    
    "🔍 Проверяем мелкие детали…",
    "🎯 Финальная корректировка результата…",
    "🌟 Последние штрихи…",    
    "🪄 Почти готово…",
        
]:
        yield (
            None,
            msg,
            gr.update(visible=False),
            gr.update(value=""),
            gr.update(value=_rating_notice_idle_html()),
            gr.update(value=_comment_notice_idle_html()),
        )
        time.sleep(2.3)

    ok, msg = allow_call(2.5)
    if not ok:
        yield (
            None,
            msg,
            gr.update(visible=False),
            gr.update(value=""),
            gr.update(value=_rating_notice_idle_html()),
            gr.update(value=_comment_notice_idle_html()),
        )
        return

    if person_pil is None:
        yield (
            None,
            "❌ Загрузите фото человека",
            gr.update(visible=False),
            gr.update(value=""),
            gr.update(value=_rating_notice_idle_html()),
            gr.update(value=_comment_notice_idle_html()),
        )
        return

    if not selected_filename:
        yield (
            None,
            "❌ Выберите одежду (клик по превью)",
            gr.update(visible=False),
            gr.update(value=""),
            gr.update(value=_rating_notice_idle_html()),
            gr.update(value=_comment_notice_idle_html()),
        )
        return

    garm = load_garment_pil(selected_filename)
    if garm is None:
        yield (
            None,
            "❌ Не удалось загрузить выбранную одежду",
            gr.update(visible=False),
            gr.update(value=""),
            gr.update(value=_rating_notice_idle_html()),
            gr.update(value=_comment_notice_idle_html()),
        )
        return

    try:
        out_img = start_tryon(human_pil=person_pil, garm_img=garm)
        yield (
            out_img,
            "✅ Готово — оцените результат и при желании оставьте комментарий",
            gr.update(visible=True),
            gr.update(value=""),
            gr.update(value=_rating_notice_idle_html()),
            gr.update(value=_comment_notice_idle_html()),
        )
    except Exception as e:
        yield (
            None,
            f"❌ Ошибка: {type(e).__name__}: {str(e)[:220]}",
            gr.update(visible=False),
            gr.update(value=""),
            gr.update(value=_rating_notice_idle_html()),
            gr.update(value=_comment_notice_idle_html()),
        )


# =========================
# Boot
# =========================
ensure_garments_available()
_default_gender = "Женская"
_initial_files = list_garments(gender=_default_gender)
_initial_items = build_gallery_items(_initial_files)

with gr.Blocks(title="Virtual Try-On Rendez-vous", css=CUSTOM_CSS) as demo:
    gr.Markdown("# Virtual Try-On Rendez-vous")

    garment_files_state = gr.State(_initial_files)
    selected_garment_state = gr.State(None)

    with gr.Row():
        with gr.Column():
            person = gr.Image(label="Фото человека", type="pil", height=420)

            gr.Markdown("""
    ### Какое фото подойдёт

    ✔ В полный рост или по пояс  
    ✔ Одежда по фигуре  
    ✔ Стоите прямо, смотрите в камеру  
    ✔ Руки и предметы не закрывают тело  
    ✔ Хороший свет, без резких теней  
    ✔ В кадре только вы
    """)

            gr.Image(
                value=_UX_EXAMPLE_PIL,
                label="",
                show_label=False,
                interactive=False,
                height=340,
                visible=bool(_UX_EXAMPLE_PIL),
            )

            gender = gr.Radio(
                choices=["Женская", "Мужская"],
                value=_default_gender,
                label="Раздел каталога",
            )

            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, show_label=False)

        with gr.Column():
            out = gr.Image(label="Результат", type="pil", height=760)

            with gr.Column(visible=False) as feedback_box:
                gr.HTML('<div class="feedback-box">')

                gr.Markdown("### Оценка результата")
                with gr.Row():
                    like_btn = gr.Button("👍 Нравится")
                    dislike_btn = gr.Button("👎 Не нравится")
                rating_notice = gr.HTML(_rating_notice_idle_html())

                gr.Markdown("### Комментарий")
                feedback_comment = gr.Textbox(
                    label="",
                    placeholder="Напишите, что понравилось или что стоит улучшить...",
                    lines=3,
                    max_lines=6,
                    show_label=False,
                )
                submit_comment_btn = gr.Button("Отправить комментарий")
                comment_notice = gr.HTML(_comment_notice_idle_html())

                gr.HTML("</div>")

    gender.change(
        fn=refresh_catalog,
        inputs=[gender],
        outputs=[garment_gallery, garment_files_state, selected_garment_state, status, selected_label],
    )

    garment_gallery.select(
        fn=on_gallery_select,
        inputs=[garment_files_state],
        outputs=[selected_garment_state, selected_label],
    )

    run.click(
        fn=tryon_ui,
        inputs=[person, selected_garment_state],
        outputs=[out, status, feedback_box, feedback_comment, rating_notice, comment_notice],
        concurrency_limit=1,
    )

    like_btn.click(
        fn=submit_like_feedback,
        inputs=[selected_garment_state],
        outputs=[status, rating_notice],
        concurrency_limit=1,
    )

    dislike_btn.click(
        fn=submit_dislike_feedback,
        inputs=[selected_garment_state],
        outputs=[status, rating_notice],
        concurrency_limit=1,
    )

    submit_comment_btn.click(
        fn=submit_comment_feedback,
        inputs=[selected_garment_state, feedback_comment],
        outputs=[status, comment_notice, feedback_comment],
        concurrency_limit=1,
    )

demo.queue(max_size=20)

if __name__ == "__main__":
    demo.launch(
        server_name="0.0.0.0",
        server_port=7860,
        share=False,
        max_threads=4,
        show_error=True,
        show_api=False,
    )