Python Experiments

Collection of Python experiments, benchmarks, and ML tools.

📂 Projects

🎬 video_ml/

AI-powered video and image enhancement toolkit with flexible pipeline support:

Core Modules (core/):

• enhancer.py - RealESRGAN for 4x super-resolution (images & videos)
• pipeline.py - VideoPipeline for video processing
• video_utils.py - Audio extraction/merging utilities

Examples (examples/):

• main.py - Single entry point for all image and video processing
• config.toml - Configuration file for customizing processing

Key Features:

• 4x super-resolution enhancement for images and videos
• Automatic audio preservation for videos
• Configuration-driven workflow
• GPU/CPU auto-detection with configurable device selection
• ffmpeg-based video encoding for maximum compatibility

Model: Pre-trained PyTorch model (64 MB)

• RealESRGAN_x4plus.pth - 4x super-resolution enhancement

📊 benchmarks/

Performance benchmarks and comparisons:

• Pandas vs Polars on 10M rows
• Vectorized vs non-vectorized operations

🔧 utils/

Utility scripts:

• examples/ - HEIC conversion, batch renaming tools

🧪 simulation/

Statistical simulations:

• examples/ - Ensemble smoothing experiments

🚀 Quick Start

# Install system dependencies
# Ubuntu/Debian:
sudo apt-get install ffmpeg

# macOS:
brew install ffmpeg

# Install package with dependencies
pip install -e .

# Download model (optional - only needed for enhancement)
bash src/video_ml/download_models.sh

# Run processing (configure in config.toml)
python -m video_ml.examples.main

# Or import core classes directly
python -c "from video_ml.core import ImageEnhancer, VideoPipeline; print('Loaded successfully')"

⚙️ Configuration

Edit src/video_ml/examples/config.toml to customize processing:

# Device configuration
[device]
device = "auto"  # Options: "cuda", "cpu", or "auto" (auto-detect GPU)

# Video enhancement configuration
[video_processing]
input_video = "src/video_ml/examples/test_inputs/test_video.mp4"
output_video = "src/video_ml/examples/test_outputs/enhanced_video.mp4"
weights_path = "model_weights/RealESRGAN_x4plus.pth"
preserve_audio = true

# Image enhancement configuration
[image_enhancer]
input_dir = "src/video_ml/examples/test_inputs"
output_dir = "src/video_ml/examples/test_outputs"
weights_path = "model_weights/RealESRGAN_x4plus.pth"
batch_size = 1

📦 Dependencies

Core packages:

• PyTorch - Deep learning framework
• OpenCV - Computer vision
• Pandas/Polars - Data analysis
• Pillow - Image processing (including HEIF support)
• ffmpeg - Video encoding (system dependency)

See pyproject.toml for complete list.

Development dependencies

pip install -e .[dev]  # Includes pytest, black, ruff, mypy

🎯 GPU Support

For NVIDIA GPU acceleration, install CUDA-enabled PyTorch:

pip install torch torchvision --index-url https://download.pytorch.org/whl/cu118