PyLandmark

A facial landmark detector and pose estimator for python.

• The landmark detector is based on OpenFace.If you use MTCNN face detector, the 5 facial landmarks can be used to get a better initial pose, which will be much faster than the original code.
• The pose estimator is from sdm.

pyboostcvconverter is used for data structure convertion between cv::Mat and numpy.

The code depends on OpenCV, other depends like OpenBlas and TBB were removed to make it easier to compile.

Depends

• CMake
• Opencv
• numpy

Setup

git clone https://github.com/cnzeki/PyLandmark
cd PyLandmark
python setup.py build
python setup.py install

echo PyLandmark model dir: `pwd`/model

Usage

# Init specify model dir
LandmarkDetector.create("./model/")

# Detect 68 landmarks
# rect:[x,y,w,h]
# point5:[x1,y2,..,x5,y5] or []
# type: 0 if point5 else 1
# return:[x1,y2,..,x68,y68]
pts = LandmarkDetector.detect(im, rect, point5, 0)

# Pose estimate
# return [pitch, yaw, roll] in degree
pose = LandmarkDetector.getPose(pts)

# Cleanup
LandmarkDetector.destroy()

Demo

Since we have to do face detection first, check out MtcnnPyCaffe alongside this project.

python demo/demo.py demo/1.jpg
...
python demo/demo.py demo/4.jpg

Check the result images in the demo directory.