search

DeepBaksuVision / YOLO9000

1 stars 3 forks source link

Darknet19 pretrain weight transfer learning #15

Open visionNoob opened 5 years ago

visionNoob commented 5 years ago
  1. Darknet19 weight 가져오기
  2. Post hoc transfer learning에 대해서 데모
visionNoob commented 5 years ago

darknet cfg 가져올때 쓸만한 https://github.com/marvis/pytorch-caffe-darknet-convert/blob/master/cfg.py

visionNoob commented 5 years ago

1: MaxPool2d 2: Sequential - conv2d, bn2, leaky2 64 3: MaxPool2d 4: Sequential - conv3d, bn3, leaky3 128 5: Sequential - conv4d, bn4, leaky4 64 6: Sequential - conv5d, bn5, leaky5 128 7: MaxPool2d 8: Sequential - conv6d, bn6, leaky6 256 9: Sequential - conv7d, bn7, leaky7 128 10: Sequential - conv8d, bn8, leaky8 256 11: MaxPool2d 12: Sequential - conv9d, bn9, leaky9 512 13: Sequential - conv10d, bn10, leaky10 256 14: Sequential - conv11d, bn11, leaky11 512 15: Sequential - conv12d, bn12, leaky12 256 16: Sequential - conv13d, bn13, leaky13 512 17: MaxPool2d 18: Sequential - conv14d, bn14, leaky14 1024 19: Sequential - conv15d, bn15, leaky15 512 20: Sequential - conv16d, bn16, leaky16 1024 21: Sequential - conv17d, bn17, leaky17 512 22: Sequential - conv18d, bn18, leaky18 1024 23: Sequential - conv19d 1000 24: GlobalAvgPool2d 25: Softmax())