Free, MIT licensed, no account required. Get the complete model package and start detecting faces immediately.
Pre-trained weights, Python implementation, usage examples, and complete documentation
✓ Clean & safe · ✓ No adware · ✓ No sign-up · ✓ MIT Licensed
MobileNet and ResNet-50 backbone weights trained on WIDER FACE, ready for immediate inference.
Complete Python implementation with inference examples, alignment utilities, and a simple CLI tool.
Step-by-step installation guide, API reference, and annotated example notebooks.
Default configuration files for both backbone variants with recommended anchor settings.
Reproduction scripts and results tables for verifying performance on WIDER FACE subsets.
Full license text confirming free use for personal and commercial projects without restriction.
Click the download button above, save the ZIP file, and extract it to your project directory.
Run pip install retinaface in your terminal to install all required Python packages.
Import RetinaFace in Python, call detect_faces(), and receive bounding boxes and landmarks instantly.