title	emoji	colorFrom	colorTo	sdk	sdk_version	app_file	pinned
Loomi Clothing Detection API	🏢	indigo	red	docker	1.0.0	main.py	false

Loomi Clothing Detection API 🚀

AI-powered clothing analysis and segmentation API, optimized for Hugging Face Spaces.

⚠️ Important Note for Hugging Face Demo

This is currently a demo version on Hugging Face Spaces with limitations:

• Sequential processing - Multiple requests are accepted but queue up and wait for the model to finish the previous request
• If you see "offline" status → wait until current request completes
• Requests will be processed in order - your request will be queued and processed when ready
• For production use → deploy to your own server with higher concurrency

✨ Features

• 🧠 AI-Powered: Uses Segformer model for clothing detection
• 🖼️ Image Processing: Background removal and dominant color detection
• ⚡ Fast: Optimized for single-request processing with automatic caching
• 🔧 HF Optimized: Built specifically for Hugging Face Spaces
• 📱 Smart Compression: WebP format with PNG fallback for optimal file sizes
• 🎯 Efficient Workflow: Two-step process for maximum performance

🚀 Quick Start

API Endpoints

• GET / - API overview and documentation
• GET /health - System health and status
• GET /performance - Performance statistics and cache info
• POST /detect - Detect clothing types with segmentation data
• POST /analyze - Analyze clothing using segmentation data (fast, no re-upload)

Usage Example

import requests

# Step 1: Upload image for clothing detection
with open('image.jpg', 'rb') as f:
    response = requests.post(
        'https://your-hf-space.hf.space/detect',
        files={'file': f}
    )
    result = response.json()
    print(result)
    
# Step 2: Analyze using segmentation data (much faster!)
analyze_response = requests.post(
    'https://your-hf-space.hf.space/analyze',
    json={
        'segmentation_data': result['segmentation_data'],
        'selected_clothing': 'shirt'  # Optional: specify clothing type
    }
)
analysis = analyze_response.json()
print(analysis)

🏗️ Architecture

• FastAPI: Modern, fast web framework
• Efficient Processing: Optimized for single requests with smart caching
• Model Management: Efficient ML model loading
• Automatic Caching: Smart caching for repeated images and segmentation data
• Image Optimization: WebP compression with PNG fallback for optimal file sizes

🔧 Configuration

The API automatically detects Hugging Face Spaces and applies optimizations:

• Single worker process
• Optimized cache sizes
• HF-specific environment variables
• Smart image compression (WebP/PNG)

📱 Integration

Perfect for:

• Mobile apps (React Native, Flutter)
• Web applications
• E-commerce platforms
• Fashion analysis tools

🚀 Running the API

# Simple startup
python run.py

# Or with uvicorn directly
uvicorn main:app --host 0.0.0.0 --port 7860

🤝 Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Submit a pull request

📚 Models and Technologies

Core ML Model:

• Segformer B2 for Clothing Segmentation - mattmdjaga/segformer_b2_clothes
  • Pre-trained model specifically designed for clothing and fashion item detection
  • Based on the Segformer architecture for semantic segmentation

Key Libraries and Licenses:

• Transformers (Hugging Face) - Apache 2.0 License
• PyTorch - BSD License
• Segformer - MIT License
• Rembg - MIT License (background removal)
• FastAPI - MIT License
• Pillow (PIL) - HPND License
• NumPy - BSD License
• scikit-learn - BSD License

Model Capabilities:

The model can detect and segment 18 different categories:

• Background, Hat, Hair, Sunglasses, Upper-clothes, Skirt, Pants, Dress, Belt
• Left/Right-shoe, Face, Left/Right-leg, Left/Right-arm, Bag, Scarf

Image Optimization:

• WebP Format: Primary format with excellent compression (70-85% smaller than PNG)
• PNG Fallback: Optimized PNG with maximum compression for compatibility
• Smart Resizing: Automatic optimization for large images
• Quality Preserved: Visual quality maintained while reducing file sizes

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

Note: While this project is MIT licensed, it uses pre-trained models and libraries with their own licenses. Please ensure compliance with all respective licenses when using this API.

---

Made with ❤️ by the Loomi Team

AI-powered clothing analysis, simplified and ready! 🎯