tensorflow face deteector

The face detection model data from wider-face，base on Faster R-CNN

Model performance

Speed: based on nvidia GTX1070ti 12G memory per second to reach about 30fps

Instructions for use:

Baidu cloud download already trained frozen_inference_graph.pb model (about 180 m), the link: https://pan.baidu.com/s/1JyT8wbH4cjzE5LBSqsMdZQ extract code:3dt5, replacement model files directoryfrozen_inference_graph.pb
run usb_cam_face.py file

Environment description:

Tensorflow-gpu ==1.10.0 opencv==3.4.2 (python). Since this model is based on the training of tensorflow detection API, the correct operation cannot be guaranteed if the tensorflow detection API is not configured

The running effect is as follows: