https://blog.csdn.net/On_theway10/article/details/102628169

• run > python demo.py

• download WIDER FACE (passcode:lsl3) face detection data then store it into ./data_set/face_detection
  • run > python ./anno_store/tool/format/transform.py change .mat(wider_face_train.mat) into .txt(anno_train.txt)
• download CNN_FacePoint face detection and landmark data then store it into ./data_set/face_landmark

• preparing data for P-Net

  • run > python detlib/preprocess/gen_Pnet_train_data.py
  • run > python detlib/preprocess/gen_landmark.py # modify the img_size = 12
  • run > python detlib/preprocess/assemble_pnet_imglist.py

• train P-Net

  • run > python detlib/train_net/train_p_net.py

• preparing data for R-Net

  • run > python detlib/preprocess/gen_Rnet_train_data.py (maybe you should change the pnet model path)
  • run > python detlib/preprocess/gen_landmark.py # modify the img_size = 24
  • run > python detlib/preprocess/assemble_rnet_imglist.py

• train R-Net

  • run > python detlib/train_net/train_r_net.py

• preparing data for O-Net

  • run > python detlib/preprocess/gen_Onet_train_data.py
  • run > python detlib/preprocess/gen_landmark.py # modify the img_size = 48
  • run > python detlib/preprocess/assemble_onet_imglist.py

• train O-Net

  • run > python detlib/train_net/train_o_net.py