face

An experiment with OpenCV 3.x face detection and recognition. Note that you will need to install the corresponding opencv_contrib modules for the face recognizer to work.

Face detection

detect.py contains sample on running face detection on webcam attached to the computer. The levelFace function tries to rotate the image so that the face detected is level (based on the eye positions).

Face recognizer training

train.py trains the OpenCV face recognizer by extracting faces from images provided under a given folder. Images for each individual should be organized in corresponding sub-folders with the folder name used by face recognizer as the labels. e.g.:

• imgdb/Barack Obama/image1.jpg
• imgdb/Barack Obama/image2.jpg
• ...
• imgdb/Donald Trump/anotherimage.png
• imgdb/Donald Trump/yetanotherimage.jpg
• ...
• imgdb/Justin Trudeau/faces.jpg
• ...

Note that each image can contain multiple faces of the same person.

Face recognition

recognize.py puts everything together. It demonstrates on training the face recognizer and feeding webcam images to recognize faces found.

Note that in order to speed up the process, the training result should run once and saved. Subsequent running of the program can load the result instead of training the recognizer again.