search

gpleiss / efficient_densenet_pytorch

A memory-efficient implementation of DenseNets
MIT License
1.52k stars 327 forks source link

Can't not train when using a 256*256 dataset #42

Closed nessieyang closed 6 years ago

nessieyang commented 6 years ago

I can run the network by using CIFAR-10 as my dataset. However, when I use my own dataset which has the size of 256256 , it can not work. I tried to transform my data into 3232, it also works. So how can I solve the scaling problem?

gpleiss commented 6 years ago

Set small_inputs=False in the model constructor.

gpleiss commented 6 years ago

small_inputs=True expects 32 x 32 images. small_input=False expects 224 x 224.