Initial commit: FASHN VTON v1.5 HuggingFace Space

Gradio demo for the open-source virtual try-on model featuring:
- Dual image input (person + garment) with category selection
- Example images for quick testing
- Direct integration with fashn-ai/fashn-vton-1.5 weights
- Discoverability tags for HuggingFace search

.gitattributes

@@ -0,0 +1,39 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text
+*.webp filter=lfs diff=lfs merge=lfs -text
+*.png filter=lfs diff=lfs merge=lfs -text
+*.jpg filter=lfs diff=lfs merge=lfs -text
+*.jpeg filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -0,0 +1,23 @@
+# Virtual environment
+.venv/
+venv/
+env/
+
+# Model weights (downloaded at runtime)
+weights/
+
+# Python
+__pycache__/
+*.py[cod]
+*$py.class
+*.so
+
+# IDE
+.idea/
+.vscode/
+*.swp
+*.swo
+
+# OS
+.DS_Store
+Thumbs.db

README.md

@@ -0,0 +1,18 @@
+---
+title: FASHN VTON v1.5
+emoji: "\U0001F457"
+colorFrom: purple
+colorTo: pink
+sdk: gradio
+sdk_version: 6.3.0
+python_version: 3.10.14
+app_file: app.py
+pinned: false
+license: apache-2.0
+short_description: Maskless try-on directly in pixel space
+tags:
+  - virtual-try-on
+  - diffusion
+  - fashion
+  - image-to-image
+---

app.py

@@ -0,0 +1,296 @@
+"""FASHN VTON v1.5 HuggingFace Space Demo."""
+
+import os
+
+import gradio as gr
+import torch
+from huggingface_hub import hf_hub_download
+from PIL import Image
+
+# Handle spaces.GPU decorator for local vs HuggingFace execution
+try:
+    import spaces
+
+    GPU_DECORATOR = spaces.GPU
+except ImportError:
+    GPU_DECORATOR = lambda func: func
+
+# ----------------- CONFIG ----------------- #
+
+SCRIPT_DIR = os.path.dirname(os.path.abspath(__file__))
+ASSETS_DIR = os.path.join(SCRIPT_DIR, "assets")
+WEIGHTS_DIR = os.path.join(SCRIPT_DIR, "weights")
+
+CATEGORIES = ["tops", "bottoms", "one-pieces"]
+GARMENT_PHOTO_TYPES = ["model", "flat-lay"]
+
+# Global pipeline instance (lazy loaded)
+_pipeline = None
+
+
+# ----------------- HELPERS ----------------- #
+
+
+def download_weights():
+    """Download model weights from HuggingFace Hub."""
+    os.makedirs(WEIGHTS_DIR, exist_ok=True)
+    dwpose_dir = os.path.join(WEIGHTS_DIR, "dwpose")
+    os.makedirs(dwpose_dir, exist_ok=True)
+
+    # Download TryOnModel weights
+    tryon_path = os.path.join(WEIGHTS_DIR, "model.safetensors")
+    if not os.path.exists(tryon_path):
+        print("Downloading TryOnModel weights...")
+        hf_hub_download(
+            repo_id="fashn-ai/fashn-vton-1.5",
+            filename="model.safetensors",
+            local_dir=WEIGHTS_DIR,
+        )
+
+    # Download DWPose models
+    dwpose_files = ["yolox_l.onnx", "dw-ll_ucoco_384.onnx"]
+    for filename in dwpose_files:
+        filepath = os.path.join(dwpose_dir, filename)
+        if not os.path.exists(filepath):
+            print(f"Downloading DWPose/{filename}...")
+            hf_hub_download(
+                repo_id="fashn-ai/DWPose",
+                filename=filename,
+                local_dir=dwpose_dir,
+            )
+
+    print("Weights downloaded successfully!")
+
+
+# ----------------- MODEL LOADING ----------------- #
+
+
+def get_pipeline():
+    """Lazy-load the pipeline on first use (ensures GPU is available on ZeroGPU)."""
+    global _pipeline
+    if _pipeline is None:
+        # Check CUDA availability (will be true inside @spaces.GPU context)
+        if not torch.cuda.is_available():
+            raise gr.Error(
+                "CUDA is not available. This demo requires a GPU to run. "
+                "If you're on HuggingFace Spaces, please try again in a moment."
+            )
+
+        # Enable TF32 for faster computation on Ampere+ GPUs
+        if torch.cuda.get_device_properties(0).major >= 8:
+            torch.backends.cuda.matmul.allow_tf32 = True
+            torch.backends.cudnn.allow_tf32 = True
+
+        print("Downloading weights (if needed)...")
+        download_weights()
+
+        print("Loading pipeline...")
+        from fashn_vton import TryOnPipeline
+
+        _pipeline = TryOnPipeline(weights_dir=WEIGHTS_DIR, device="cuda")
+        print("Pipeline loaded on CUDA!")
+
+    return _pipeline
+
+
+# ----------------- INFERENCE ----------------- #
+
+
+@GPU_DECORATOR
+def try_on(
+    person_image: Image.Image,
+    garment_image: Image.Image,
+    category: str,
+    garment_photo_type: str,
+    num_timesteps: int,
+    guidance_scale: float,
+    seed: int,
+    segmentation_free: bool,
+) -> Image.Image:
+    """Run virtual try-on inference."""
+    if person_image is None:
+        raise gr.Error("Please upload a person image")
+    if garment_image is None:
+        raise gr.Error("Please upload a garment image")
+
+    # Handle seed (guard against None or invalid values)
+    if seed is None or seed < 0:
+        seed = 42
+
+    # Convert to RGB if needed
+    if person_image.mode != "RGB":
+        person_image = person_image.convert("RGB")
+    if garment_image.mode != "RGB":
+        garment_image = garment_image.convert("RGB")
+
+    # Get pipeline (lazy loads on first call)
+    pipeline = get_pipeline()
+
+    # Run inference
+    result = pipeline(
+        person_image=person_image,
+        garment_image=garment_image,
+        category=category,
+        garment_photo_type=garment_photo_type,
+        num_samples=1,
+        num_timesteps=num_timesteps,
+        guidance_scale=guidance_scale,
+        seed=int(seed),
+        segmentation_free=segmentation_free,
+    )
+
+    return result.images[0]
+
+
+# ----------------- UI ----------------- #
+
+# Custom CSS
+CUSTOM_CSS = """
+.contain img {
+    object-fit: contain !important;
+    max-height: 856px !important;
+    max-width: 576px !important;
+}
+"""
+
+# Load HTML content
+with open(os.path.join(SCRIPT_DIR, "banner.html"), "r") as f:
+    banner_html = f.read()
+with open(os.path.join(SCRIPT_DIR, "tips.html"), "r") as f:
+    tips_html = f.read()
+
+# Build example paths
+examples_dir = os.path.join(ASSETS_DIR, "examples")
+
+# Paired examples: [person_path, garment_path, category, garment_photo_type]
+paired_examples = [
+    [os.path.join(examples_dir, "person1.png"), os.path.join(examples_dir, "garment1.jpeg"), "one-pieces", "model"],
+    [os.path.join(examples_dir, "person2.png"), os.path.join(examples_dir, "garment2.webp"), "tops", "model"],
+    [os.path.join(examples_dir, "person3.png"), os.path.join(examples_dir, "garment3.jpeg"), "tops", "flat-lay"],
+    [os.path.join(examples_dir, "person4.png"), os.path.join(examples_dir, "garment4.webp"), "tops", "model"],
+    [os.path.join(examples_dir, "person5.png"), os.path.join(examples_dir, "garment5.jpeg"), "bottoms", "flat-lay"],
+    [os.path.join(examples_dir, "person6.png"), os.path.join(examples_dir, "garment6.webp"), "one-pieces", "model"],
+]
+
+# Individual examples (classic from repo)
+person_only_examples = [os.path.join(examples_dir, "person0.png")]
+garment_only_examples = [os.path.join(examples_dir, "garment0.png")]
+
+# Build UI
+with gr.Blocks(css=CUSTOM_CSS) as demo:
+    # Header
+    gr.HTML(banner_html)
+    gr.HTML(tips_html)
+
+    with gr.Row(equal_height=False):
+        # Left column: Inputs
+        with gr.Column(scale=1):
+            person_image = gr.Image(
+                label="Person Image",
+                type="pil",
+                sources=["upload", "clipboard"],
+                elem_classes=["contain"],
+            )
+
+            # Individual person examples
+            gr.Examples(
+                examples=person_only_examples,
+                inputs=person_image,
+                label="Person Examples",
+            )
+
+            garment_image = gr.Image(
+                label="Garment Image",
+                type="pil",
+                sources=["upload", "clipboard"],
+                elem_classes=["contain"],
+            )
+
+            # Individual garment examples
+            gr.Examples(
+                examples=garment_only_examples,
+                inputs=garment_image,
+                label="Garment Examples",
+            )
+
+            with gr.Row():
+                category = gr.Dropdown(
+                    choices=CATEGORIES,
+                    value="tops",
+                    label="Garment Category",
+                )
+                garment_photo_type = gr.Dropdown(
+                    choices=GARMENT_PHOTO_TYPES,
+                    value="model",
+                    label="Garment Photo Type",
+                )
+
+            run_button = gr.Button("Try On", variant="primary", size="lg")
+
+            # Advanced settings
+            with gr.Accordion("Advanced Settings", open=False):
+                num_timesteps = gr.Slider(
+                    minimum=10,
+                    maximum=50,
+                    value=50,
+                    step=5,
+                    label="Sampling Steps",
+                    info="Higher = better quality, slower.",
+                )
+                guidance_scale = gr.Slider(
+                    minimum=1.0,
+                    maximum=3.0,
+                    value=1.5,
+                    step=0.1,
+                    label="Guidance Scale",
+                    info="How closely to follow the garment. 1.5 recommended.",
+                )
+                seed = gr.Number(
+                    value=42,
+                    label="Seed",
+                    info="Random seed for reproducibility.",
+                    precision=0,
+                )
+                segmentation_free = gr.Checkbox(
+                    value=True,
+                    label="Segmentation Free",
+                    info="Preserves body features and allows unconstrained garment volume. Disable for tighter garment fitting.",
+                )
+
+        # Right column: Output
+        with gr.Column(scale=1):
+            result_image = gr.Image(
+                label="Try-On Result",
+                type="pil",
+                interactive=False,
+                elem_classes=["contain"],
+            )
+
+    # Paired examples at the bottom
+    gr.Examples(
+        examples=paired_examples,
+        inputs=[person_image, garment_image, category, garment_photo_type],
+        label="Complete Examples (click to load person + garment + settings)",
+    )
+
+    # Event handler
+    run_button.click(
+        fn=try_on,
+        inputs=[
+            person_image,
+            garment_image,
+            category,
+            garment_photo_type,
+            num_timesteps,
+            guidance_scale,
+            seed,
+            segmentation_free,
+        ],
+        outputs=[result_image],
+    )
+
+# Configure queue with concurrency limit to prevent GPU OOM
+demo.queue(default_concurrency_limit=1, max_size=30)
+
+if __name__ == "__main__":
+    demo.launch(share=False)

banner.html

@@ -0,0 +1,59 @@
+<div style="
+      display: flex;
+      flex-direction: column;
+      justify-content: center;
+      align-items: center;
+      text-align: center;
+    background: linear-gradient(45deg, #1a1a1a 0%, #333333 100%);
+      padding: 24px;
+      gap: 24px;
+      border-radius: 8px;
+    ">
+  <div style="display: flex; gap: 8px;">
+    <h1 style="
+  font-size: 42px;
+  color: #fafafa;
+  margin: 0;
+  font-family: 'Trebuchet MS', 'Lucida Sans Unicode', 'Lucida Grande',
+    'Lucida Sans', Arial, sans-serif;
+">
+      FASHN VTON v1.5
+    </h1>
+  </div>
+
+  <p style="
+          margin: 0;
+          line-height: 1.6rem;
+          font-size: 18px;
+          color: #fafafa;
+          opacity: 0.9;
+        ">
+    Efficient Maskless Virtual Try-On in Pixel Space
+  </p>
+
+  <p style="
+          margin: 0;
+          line-height: 1.5rem;
+          font-size: 14px;
+          color: #fafafa;
+          opacity: 0.7;
+        ">
+    Virtual try-on model that generates photorealistic images directly in pixel space without requiring segmentation masks.
+  </p>
+
+  <div style="
+        display: flex;
+        justify-content: center;
+        align-items: center;
+        text-align: center;
+        gap: 4px;
+        flex-wrap: wrap;
+      ">
+    <a href="https://fashn.ai"><img src="https://custom-icon-badges.demolab.com/badge/FASHN_AI-333333?style=for-the-badge&logo=fashn" alt="FASHN AI" /></a>
+    <a href="https://fashn.ai/research/vton-1-5"><img src="https://img.shields.io/badge/Project-Page-1A1A1A?style=for-the-badge" alt="Project Page" /></a>
+    <a href="https://github.com/fashn-AI/fashn-vton-1.5"><img src="https://img.shields.io/badge/github-%23121011.svg?style=for-the-badge&logo=github&logoColor=white" alt="Github" /></a>
+    <a href="https://huggingface.co/fashn-ai/fashn-vton-1.5"><img src="https://img.shields.io/badge/%F0%9F%A4%97%20Model-FFD21E?style=for-the-badge" alt="HuggingFace Model" /></a>
+    <a href="#"><img src="https://img.shields.io/badge/arXiv-Coming%20Soon-b31b1b?style=for-the-badge&logo=arXiv&logoColor=white" alt="arXiv" /></a>
+    <a href="https://discord.gg/4XRXgJ2ysU"><img src="https://img.shields.io/badge/Discord-%235865F2.svg?style=for-the-badge&logo=discord&logoColor=white" alt="Discord" /></a>
+  </div>
+</div>

requirements.txt

@@ -0,0 +1,10 @@
+--extra-index-url https://download.pytorch.org/whl/cu121
+torch
+torchvision
+gradio==6.3.0
+fashn-vton @ git+https://github.com/fashn-AI/fashn-vton-1.5.git@38aafe2185df40a3e8a5442e950c422c3d9dcb5a
+onnxruntime-gpu>=1.14.0
+spaces
+pillow
+numpy
+huggingface_hub

tips.html

@@ -0,0 +1,29 @@
+<div style="
+    padding: 12px;
+    border: 1px solid #333333;
+    border-radius: 8px;
+    text-align: center;
+    display: flex;
+    flex-direction: column;
+    gap: 8px;
+  ">
+  <b style="font-size: 18px;">Tips for best results</b>
+
+  <ul style="
+      display: flex;
+      gap: 16px;
+      justify-content: center;
+      list-style: none;
+      padding: 0;
+      margin: 0;
+      flex-wrap: wrap;
+    ">
+    <li>Single person, clearly visible</li>
+    <li>|</li>
+    <li>Match category to garment type</li>
+    <li>|</li>
+    <li>Use "flat-lay" for product shots</li>
+    <li>|</li>
+    <li>2:3 aspect ratio optimal</li>
+  </ul>
+</div>