szad670401/HyperLandmark
A CNN-based face landmark detector that locates 106 facial keypoints in real time on Android mobile devices.
Not currently ranked — collecting fresh signals.
star history
HyperLandmark provides a deep learning-based solution for mobile face landmark detection, identifying 106 facial keypoints for applications like face alignment, beautification, and liveness detection. The Android implementation uses a compact CNN model (~2MB) optimized for mobile inference, achieving around 7ms per face on Qualcomm 820 hardware. The project includes multi-face tracking support and serves as a preprocessing step for more complex facial analysis pipelines.
Frequently asked
- What is szad670401/HyperLandmark?
- A CNN-based face landmark detector that locates 106 facial keypoints in real time on Android mobile devices.
- Is HyperLandmark open source?
- Yes — szad670401/HyperLandmark is open source, released under the Apache-2.0 license.
- What language is HyperLandmark written in?
- szad670401/HyperLandmark is primarily written in C++.
- How popular is HyperLandmark?
- szad670401/HyperLandmark has 1.7k stars on GitHub.
- Where can I find HyperLandmark?
- szad670401/HyperLandmark is on GitHub at https://github.com/szad670401/HyperLandmark.