CNN based face detection from dlib http://www.arunponnusamy.com/cnn-face-detector-dlib.html

Abstraction: Comparing dlib HOG and CNN face detectors in Python with timing

Key points:

• Demonstrates both dlib.get_frontal_face_detector() (HOG+SVM) and dlib.cnn_face_detection_model_v1() (CNN) in one script
• CNN detector requires a weights file (mmod_human_face_detector.dat); HOG detector is built-in with no extra files
• Both detectors are run and timed separately; execution times printed to compare speed tradeoff
• HOG results drawn in green; CNN results drawn in red — visual differentiation on the same output image
• CNN detector returns face.rect objects (not direct rectangles), requiring .rect.left()/.top()/.right()/.bottom() accessors

Connections: Dlib · Face Detection · Convolutional Neural Networks

Source: https://gist.github.com/arunponnusamy/7013c8617f97937250cb2c2de57c9b11