All notable changes to PrimateFace will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
-
Core Framework
- Cross-species primate face detection using mmdetection
- 68-point facial landmark estimation with mmpose
- Support for multiple primate species (macaques, marmosets, lemurs, chimpanzees, etc.)
- COCO format annotations for standardized data handling
-
Model Support
- Integration with MMDetection and MMPose frameworks
- Ultralytics YOLO support for real-time inference
- DeepLabCut and SLEAP integration for behavioral analysis
- Pre-trained models optimized for primate faces
-
Tools & Features
- Interactive GUI for pseudo-labeling and annotation refinement
- DINOv2 feature extraction for unsupervised analysis
- Landmark converter for 68→49 point format conversion
- Parallel GPU processing for large-scale video analysis
- Temporal smoothing for stable video tracking
-
Documentation & Examples
- 6 interactive Jupyter notebook tutorials
- Comprehensive API documentation
- Installation guides for multiple platforms
- Example scripts for common workflows
- VLM genus classification requires >PyTorch 2.1.0 specifically
- High GPU memory usage for batch processing (recommend 8GB+ VRAM)
- Enhanced cross-species generalization
- Real-time video processing optimizations
- Additional pre-trained models for rare species
- Docker container for easy deployment
- Extended dataset with additional annotations
- Monocular 3D facial reconstruction from 2D landmarks
- Custom cross-species gaze tracking
- Automated pipelines for facial action unit recognition
- Integration with additional frameworks
See CONTRIBUTING.md for information on how to contribute to PrimateFace.