Update README.md

README.md

@@ -8,294 +8,93 @@ sdk_version: "4.44.1"
 app_file: app.py
 pinned: false
 license: mit
-python_version: "3.9"
----
-
-# Video Background Replacement
-
-Professional-quality video background replacement using SAM2 + MatAnyone for precise segmentation and cinema-grade compositing.
-
-## Features
-
-### Core Capabilities
-- **High-Quality Segmentation**: SAM2-powered person detection with sub-pixel accuracy
-- **Advanced Matting**: MatAnyone integration for professional edge refinement
-- **Dual Processing Modes**:
-  - **Single-Stage**: Direct background replacement (fast)
-  - **Two-Stage**: Green screen intermediate for broadcast-quality results
-- **Multiple Background Options**: Custom uploads, professional presets, procedural generation
-- **GPU Accelerated**: CUDA optimization for NVIDIA GPUs with CPU fallback
-
-### Technical Highlights
-- Keyframe-based processing with temporal consistency
-- Automatic memory management and error recovery
-- Real-time progress tracking with ETA estimation
-- Audio preservation throughout processing
-- Robust codec fallback system
-
-## Quick Start
-
-### Option 1: Docker (Recommended)
-
-```bash
-# Clone repository
-git clone <your-repo-url>
-cd video-background-replacement
-
-# Build and run with GPU support
-docker build -t video-bg-replacement .
-docker run --gpus all -p 7860:7860 video-bg-replacement
-```
-
-### Option 2: Local Installation
-
-```bash
-# Clone repository
-git clone <your-repo-url>
-cd video-background-replacement
-
-# Create virtual environment
-python -m venv venv
-source venv/bin/activate  # On Windows: venv\Scripts\activate
-
-# Install dependencies
-pip install -r requirements.txt
-
-# Run application
-python app.py
-```
-
-## System Requirements
-
-### Minimum Requirements
-- Python 3.10+
-- 8GB RAM
-- 4GB storage space
-- FFmpeg installed
-
-### Recommended for Best Performance
-- NVIDIA GPU with 6GB+ VRAM
-- 16GB+ RAM
-- CUDA 12.1+ support
-- Fast SSD storage
-
-### Supported Platforms
-- Linux (Ubuntu 20.04+, tested)
-- Windows 10/11 with WSL2
-- macOS (CPU-only, limited testing)
-
-## Usage Guide
-
-### Basic Workflow
-
-1. **Launch Application**
-   ```bash
-   python app.py
-   ```
-   Access web interface at `http://localhost:7860`
-
-2. **Load Models** (first-time setup)
-   - Click "Load Models" button
-   - Wait for SAM2 and MatAnyone to download and initialize
-   - Status will show "Models loaded and validated"
-
-3. **Process Video**
-   - Upload your video file (MP4, AVI, MOV supported)
-   - Choose background method:
-     - **Professional Presets**: Studio-quality backgrounds
-     - **Custom Upload**: Your own background image
-   - Select processing options:
-     - **Two-Stage Mode**: Better quality, slower processing
-     - **Quality Preset**: Fast/Balanced/High
-   - Click "Process Video"
-
-### Processing Modes
-
-#### Single-Stage Mode (Default)
-- Direct background replacement
-- Faster processing (2-5x speed)
-- Good quality for most use cases
-- Recommended for: Social media, quick edits, testing
-
-#### Two-Stage Mode (Premium)
-- Green screen intermediate step
-- Cinema-quality edge compositing
-- Advanced chroma key algorithms
-- Recommended for: Professional content, broadcast, film
-
-### Background Options
-
-#### Professional Presets
-- `office_modern`: Clean contemporary office
-- `studio_blue`: Broadcast-quality blue background
-- `studio_green`: Professional green screen replacement
-- `minimalist`: Clean white gradient
-- `warm_gradient`: Warm sunset atmosphere
-- `tech_dark`: Modern tech/gaming setup
-
-#### Custom Backgrounds
-- Upload any image (JPG, PNG supported)
-- Automatically resized to match video resolution
-- Best results with high-resolution images (1920x1080+)
-
-## Configuration
-
-### Environment Variables
-
-```bash
-# Model settings
-export MODEL_CACHE_DIR="/path/to/model/cache"
-export FORCE_CPU="false"
-export DISABLE_MATANYONE="false"
-
-# Processing settings
-export KEYFRAME_INTERVAL="5"
-export FRAME_SKIP="1"
-export QUALITY_PRESET="balanced"
-
-# Video encoding
-export OUTPUT_CODEC="mp4v"
-export CRF="18"
-```
-
-### Quality Presets
-
-| Preset | Speed | Quality | Use Case |
-|--------|-------|---------|----------|
-| `fast` | 3x faster | Good | Social media, previews |
-| `balanced` | Normal | High | General use |
-| `high` | 2x slower | Excellent | Professional content |
-
-## API Reference
-
-### Core Functions
-
-```python
-from app import process_video_fixed, load_models_with_validation
-
-# Load models
-status = load_models_with_validation()
-
-# Process video
-result_path, message = process_video_fixed(
-    video_path="input.mp4",
-    background_choice="studio_blue",
-    custom_background_path=None,
-    use_two_stage=False,
-    chroma_preset="standard"
-)
-```
-
-### Two-Stage Processing
-
-```python
-from two_stage_processor import TwoStageProcessor
-
-processor = TwoStageProcessor(sam2_predictor, matanyone_model)
-
-# Full pipeline
-result_path, message = processor.process_full_pipeline(
-    video_path="input.mp4",
-    background=background_image,
-    final_output="output.mp4",
-    chroma_settings={"tolerance": 40, "edge_softness": 2}
-)
-```
-
-## Troubleshooting
-
-### Common Issues
-
-**CUDA Out of Memory**
-```bash
-# Reduce processing quality
-export QUALITY_PRESET="fast"
-export KEYFRAME_INTERVAL="8"
-```
-
-**Models Not Loading**
-```bash
-# Clear cache and retry
-rm -rf /tmp/model_cache
-python app.py
-```
-
-**Video Processing Fails**
-- Check video format (MP4 recommended)
-- Ensure video is not corrupted
-- Try shorter clips first (under 30 seconds)
-
-**Audio Missing**
-- FFmpeg must be installed and in PATH
-- Check input video has audio track
-- Try different output format
-
-### Performance Optimization
-
-**For Large Videos**
-- Use "fast" quality preset
-- Increase `KEYFRAME_INTERVAL` to 8-10
-- Process in shorter segments
-
-**For High Resolution**
-- Ensure sufficient VRAM (6GB+ recommended)
-- Use two-stage mode for best quality
-- Consider downscaling input video
-
-## Development
-
-### Project Structure
-```
-├── app.py                    # Main application entry point
-├── utilities.py              # Core CV functions (segmentation, compositing)
-├── two_stage_processor.py    # Green screen pipeline
-├── ui_components.py          # Gradio interface
-├── requirements.txt          # Python dependencies
-├── Dockerfile               # Container configuration
-└── README.md               # This file
-```
-
-### Contributing
-
-1. Fork the repository
-2. Create feature branch (`git checkout -b feature/amazing-feature`)
-3. Commit changes (`git commit -m 'Add amazing feature'`)
-4. Push to branch (`git push origin feature/amazing-feature`)
-5. Open Pull Request
-
-### Testing
-
-```bash
-# Run with test video
-python app.py --test-mode
-
-# Process sample
-python -c "
-from app import process_video_fixed, load_models_with_validation
-load_models_with_validation()
-result = process_video_fixed('test_video.mp4', 'office_modern', None)
-print(f'Result: {result}')
-"
-```
-
-## License
-
-MIT License - see LICENSE file for details.
-
-## Acknowledgments
-
-- **SAM2**: Meta's Segment Anything 2 for segmentation
-- **MatAnyone**: High-quality image matting
-- **Gradio**: Web interface framework
-- **OpenCV**: Computer vision processing
-- **FFmpeg**: Video encoding/decoding
-
-## Support
-
-- **Issues**: Report bugs via GitHub Issues
-- **Discussions**: Feature requests and questions
-- **Documentation**: Check troubleshooting section first
-
+python_version: "3.10"
 ---
 
+# ============================================================================
+# PYTORCH CUDA WHEELS (CUDA 12.1)
+# ============================================================================
+--extra-index-url https://download.pytorch.org/whl/cu121
+
+# ============================================================================
+# CORE PYTHON DEPENDENCIES (Python 3.10)
+# ============================================================================
+numpy==1.26.4
+Pillow>=10.0.1,<11.0
+setuptools>=65.7.0,<69.0
+wheel>=0.40.0,<1.0
+typing-extensions>=4.12.2,<5.0
+
+# ============================================================================
+# WEB FRAMEWORK & UI
+# ============================================================================
+gradio==4.44.1
+gradio_client==1.3.0
+
+# ============================================================================
+# DEEP LEARNING & AI MODELS (Python 3.10 + CUDA 12.1)
+# ============================================================================
+torch==2.1.0
+torchvision==0.16.0
+torchaudio==2.1.0
+
+# Hugging Face ecosystem - MatAnyOne compatible versions
+transformers==4.43.3
+huggingface_hub==0.24.5
+accelerate>=0.20.3,<1.0
+safetensors==0.4.3
+
+# Model utilities - MatAnyOne requirements
+einops==0.8.0
+timm>=0.9.16
+
+# ============================================================================
+# COMPUTER VISION & VIDEO PROCESSING
+# ============================================================================
+opencv-python-headless==4.10.0.84
+
+# Video processing
+moviepy>=1.0.3,<2.0
+imageio==2.34
+imageio-ffmpeg>=0.4.8,<1.0
+ffmpeg-python>=0.2.0,<1.0
+
+# ============================================================================
+# SCIENTIFIC COMPUTING
+# ============================================================================
+scipy==1.13.1
+tqdm>=4.66.1,<5.0
+
+# ============================================================================
+# CONFIGURATION & UTILITIES
+# ============================================================================
+hydra-core==1.3.2
+omegaconf==2.3.0
+diskcache>=5.6.3,<6.0
+psutil>=5.9.0,<6.0
+
+# ============================================================================
+# MATANYONE DEPENDENCIES (Python 3.10 compatible)
+# ============================================================================
+easydict==1.10
+gdown>=4.7.1
+hickle>=5.0
+cchardet>=2.1.7
+gitpython>=3.1
+netifaces>=0.11.0
+pycocotools>=2.0.7
+tensorboard>=2.11
+protobuf<4  # Uncommented to prevent potential conflicts
+
+# ============================================================================
+# GIT DEPENDENCIES (Model Repositories)
+# ============================================================================
+# SAM2 (Python 3.10+)
+git+https://github.com/facebookresearch/segment-anything-2.git@2b90b9f5ceec907a1c18123530e92e794ad901a4
+
+# MatAnyOne
+git+https://github.com/pq-yang/MatAnyOne.git@2234ce5cdc487749515518bd035b5e18bccea3da
+
+# Thin Plate Spline (MatAnyOne dependency)
+git+https://github.com/cheind/py-thin-plate-spline@f6995795397118b7d0ac01aecd3f39ffbfad9dee
 *For deployment on Hugging Face Spaces, see the space configuration in the app header.*