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.