Upload 3 files

background-removal

README.md

@@ -1,13 +1,106 @@
 ---
-title: Background Removal
-emoji: 🏃
-colorFrom: indigo
-colorTo: indigo
+title: ClosetConnect Background Removal API
+emoji: 🎨
+colorFrom: blue
+colorTo: purple
 sdk: gradio
-sdk_version: 6.3.0
+sdk_version: 4.0.0
 app_file: app.py
 pinned: false
-short_description: ' background-removal'
+license: mit
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# 🎨 Background Removal API
+
+Fast background removal service using **rembg** (u2net model) optimized for Hugging Face Spaces with GPU support.
+
+## Features
+
+- ⚡ **GPU-accelerated** processing (1-3 seconds per image)
+- 🔄 **Model session reuse** for optimal performance
+- 🌐 **REST API** endpoint for external services
+- 🎯 **Optimized for clothing images** (used in ClosetConnect project)
+
+## Usage
+
+### Web Interface
+
+Simply upload an image and get the background removed instantly.
+
+### API Endpoint
+
+You can call this service from your applications:
+
+```python
+import requests
+import base64
+from PIL import Image
+import io
+
+# Your Hugging Face Space URL
+API_URL = "https://YOUR-USERNAME-background-removal.hf.space/api/predict"
+
+# Read image
+with open("your_image.png", "rb") as f:
+    image_bytes = f.read()
+
+# Call API
+response = requests.post(
+    API_URL,
+    files={"data": ("image.png", image_bytes, "image/png")},
+    timeout=60
+)
+
+# Parse response
+result = response.json()
+if "data" in result and len(result["data"]) > 0:
+    # Extract base64 image data
+    image_data_url = result["data"][0]
+    base64_data = image_data_url.split(",")[1]
+    image_data = base64.b64decode(base64_data)
+
+    # Save result
+    output_image = Image.open(io.BytesIO(image_data))
+    output_image.save("output.png")
+```
+
+### Environment Variables for CloudRun Worker
+
+Set this in your CloudRun Worker environment:
+
+```bash
+REMBG_API_URL=https://YOUR-USERNAME-background-removal.hf.space
+```
+
+## Performance
+
+- **First request**: 5-10s (model loading)
+- **Subsequent requests**: 1-3s (GPU)
+- **Memory usage**: ~500MB (model + runtime)
+
+## Model
+
+- **Model**: u2net (176MB)
+- **Framework**: rembg 2.0.55+
+- **Inference**: ONNX Runtime with GPU support
+
+## Deployment to Hugging Face Spaces
+
+1. Create a new Space on [Hugging Face](https://huggingface.co/spaces)
+2. Choose **Gradio** SDK
+3. Enable **GPU** (Settings → Hardware → GPU)
+4. Upload these files:
+   - `app.py`
+   - `requirements.txt`
+   - `README.md`
+5. The Space will automatically build and deploy
+
+## License
+
+MIT License - Free to use for personal and commercial projects
+
+## Related
+
+- Part of the **ClosetConnect** project
+- Used for clothing image preprocessing
+- Integrated with CloudRun Worker pipeline

app.py

@@ -0,0 +1,96 @@
+"""
+Background Removal API using rembg
+Optimized for Hugging Face Spaces with GPU support
+
+This Gradio app provides a simple API endpoint for background removal
+that can be called from CloudRun Worker or other services.
+"""
+
+import gradio as gr
+from rembg import remove, new_session
+from PIL import Image
+import io
+import time
+
+# Global session for model reuse (GPU optimized)
+print("🔥 Loading rembg model...")
+start_time = time.time()
+rembg_session = new_session(model_name='u2net')
+load_time = time.time() - start_time
+print(f"✅ rembg model loaded in {load_time:.2f}s")
+
+def remove_background(image):
+    """
+    Remove background from image using rembg
+
+    Args:
+        image: PIL Image
+
+    Returns:
+        PIL Image with transparent background
+    """
+    print(f"📥 Received image: {image.size}, mode: {image.mode}")
+
+    try:
+        # Convert PIL Image to bytes
+        img_byte_arr = io.BytesIO()
+        image.save(img_byte_arr, format='PNG')
+        img_byte_arr.seek(0)
+        image_bytes = img_byte_arr.getvalue()
+
+        # Remove background using rembg with session reuse
+        print("🔄 Processing with rembg...")
+        start_time = time.time()
+        output_bytes = remove(image_bytes, session=rembg_session)
+        process_time = time.time() - start_time
+        print(f"✅ Background removed in {process_time:.2f}s")
+
+        # Convert back to PIL Image
+        output_image = Image.open(io.BytesIO(output_bytes)).convert("RGBA")
+        print(f"📤 Output image: {output_image.size}, mode: {output_image.mode}")
+
+        return output_image
+
+    except Exception as e:
+        print(f"❌ Error: {e}")
+        raise e
+
+# Create Gradio Interface
+demo = gr.Interface(
+    fn=remove_background,
+    inputs=gr.Image(type="pil", label="Upload Image"),
+    outputs=gr.Image(type="pil", label="Background Removed"),
+    title="🎨 Background Removal API",
+    description="""
+    Fast background removal using rembg (u2net model).
+
+    **Features:**
+    - GPU-accelerated processing (on HF Spaces)
+    - Session reuse for faster processing
+    - API endpoint available at `/api/predict`
+
+    **Usage:**
+    ```python
+    import requests
+
+    response = requests.post(
+        "YOUR_SPACE_URL/api/predict",
+        files={"data": open("image.png", "rb")}
+    )
+    result = response.json()
+    ```
+    """,
+    examples=[
+        # You can add example images here if you want
+    ],
+    api_name="predict",  # API endpoint name
+    allow_flagging="never"
+)
+
+if __name__ == "__main__":
+    # Launch with API mode enabled
+    demo.launch(
+        server_name="0.0.0.0",
+        server_port=7860,
+        show_api=True
+    )

requirements.txt

@@ -0,0 +1,4 @@
+gradio>=4.0.0
+rembg>=2.0.55
+pillow>=10.1.0
+onnxruntime-gpu>=1.16.0