This repo is inspired by face_net.
Use Python 3.5 or above
This repo contains two features:
- Face detection and Face alignment
- Face verification (compare) (Doing)
Face detection and Face alignment
MTCNN is used to perform face detection and alignment tasks. To run this feature on terminal, using commands below:
cd YOUR_PATH/face_detect_dojo/
python -m verification.face_detection
Or you can simply open verification/face_detection.py
file in PyCharm and run this script.
What this script does is to find test.jpg
in face_detect_dojo/resources/test_images/
and try to detect the faces (if any) in the image. So please make sure there is an image named test.jpg
in the dir.
You can see the detection and alignment result in dir face_detect_dojo/resources/debug_result/boxed_faces/
, in which an image file whose name starts with BoxedFaces
will exist.
Face verification (compare) (Doing)
This feature is still on going.
P.S.
- Visiualize MTCNN Model. By setting
DEBUG_TENSORFLOW_GRAPH = True
insettings.py
, and running Face Detection code as said above, you can visiualize MTCNN Model graph by typing commandtensorboard --logdir=face_detection_dojo/resources/event_log