search

wasidennis / AdaptSegNet

Learning to Adapt Structured Output Space for Semantic Segmentation, CVPR 2018 (spotlight)
847 stars 203 forks source link

According to the code, batch-size is 1 for trainning? #26

Closed EthanZhangYi closed 5 years ago

zhyx12 commented 5 years ago

batch-size=1 works file since the wight and bias are fix（requires_grad=False in ResNetMulti class）

EthanZhangYi commented 5 years ago

So for VGG net without Batch Normalization layer, batch_size = 1 is ok for trainning all the parameters? That is very interesting.