A pytorch implementation of the face landmarking tool from OpenPose.
https://github.com/CMU-Perceptual-Computing-Lab/openpose
Please observe the licensing terms of the original authors, which does not permit commercial use.
cd torchface
pip install -e .
On first use, the model weights should download to your home directory.
from face.pose import Face
facedetector = Face()
img = Image.open("face.jpg")
lmk = facedetector.detect(img)
It is possible to expose the heatmaps with:
heatmaps = facedetector._detect(img)
there is an example included:
python test.py