TO-DO.txt

Plan
(PyTorch)

Extract feature before the first (or second?) block

Extract the top layers from the pretrained network (directly take? or remove all previous layers? or rebuild and load parameters?)

Replace the fully-connected layer

Train fc only

Train top fc and last block

Train fc and last 2 blocks




Info
https://discuss.pytorch.org/t/weight-initilzation/157
***https://discuss.pytorch.org/t/given-vgg16-how-to-remove-pool5-layer-and-all-the-classify-layers-and-how-to-add-new-layers-to-that-while-keeping-the-pretrained-weights/1653/4