pymtcnn 1.1.1

High-performance cross-platform MTCNN face detection with CUDA and Apple Neural Engine support

pymtcnn

MTCNN face detection with CoreML (Apple Silicon) and CUDA support.

Installation

pip install pymtcnn

Usage

from pymtcnn import MTCNN

detector = MTCNN()  # auto-selects best backend
boxes, landmarks = detector.detect(image)

What it does

• Detects faces and 5-point facial landmarks
• Auto-selects backend: CoreML on Mac, CUDA on NVIDIA, CPU fallback
• ~34 FPS on Apple Silicon, ~50 FPS on CUDA

Citation

If you use this in research, please cite:

Wilson IV, J., Rosenberg, J., Gray, M. L., & Razavi, C. R. (2025). A split-face computer vision/machine learning assessment of facial paralysis using facial action units. Facial Plastic Surgery & Aesthetic Medicine. https://doi.org/10.1177/26893614251394382

License

CC BY-NC 4.0 — free for non-commercial use with attribution.