The pretraining of VGG19

Open hemp110 opened this issue 8 years ago • 3 comments

Thanks for sharing the code. When I run the BGAN code, I find that the pretrained vgg19-like model is crucial. So how to train the model on ImageNet since it has two branch layers fc8 and fc10?

Nov 09 '17 05:11 hemp110

Actually, we could only care the fc6 and fc7 layers, so fc8 and fc10 layers have no need to pre-train.

Nov 09 '17 07:11 ht014

Thanks for sharing the code. I wander how long it will take to train your BGAN.py program with CPU?

Jan 19 '18 02:01 wu4yan

I haven't tested it on CPU, but I trained it on GPU, which costed about 5 hours.

Jan 22 '18 02:01 ht014