Yubin Wang

@twmht you can generate your new prototxt based on pynetbuild.

@twmht sorry for typo. [pynetbuilder](https://github.com/jay-mahadeokar/pynetbuilder)

@minixalpha any news ?

@deeprun the VGG module is not the whole VGG16, which is removed fc layers, so the you can input any resolution to the network.

@jcjohnson have this implementation support cudnn v5? The Wingrad algorithm will benefit vgg-like model a lot.

@maolin23 @jgong5 @fyu have you solved the problem? I also met this problem, after I changed the finaly layer with 'xavier' initialization, the loss seems better. But I have not...

@jgong5 Unfortunately, the loss becomes bigger after 200 iteration with batchsize 6.

@jgong5 @fyu After I initialized the net with vgg_conv and initialization weights with xavier, the training loss seems better, and going down as the iteration increase, after 3w iteration, the...

@rogersct receptive field size can be calculate based on your network parameter such as conv kernel size, stride.

on Windows, anaconda is usually used for installing python packages.