In this repository you can try how Viola-Jones face detection has much smaller accuracy than using CNN from DLIB. I also implemented IOU function which you can use for calculating accuracy and two functions which will scrape 'data' folder and save faces data with pickle.