search

hezhangsprinter / DCPDN

Densely Connected Pyramid Dehazing Network (CVPR'2018)
404 stars 112 forks source link

The result images seem not to be enhanced. #22

Closed xiaofeng94 closed 5 years ago

xiaofeng94 commented 5 years ago

Hi @hezhangsprinter ,

I really appreciate this work and start trying to test your method with testing images (nature) and pre-trained model (netG_epoch_8.pth) you provide. I change netG.load_state_dict(torch.load(opt.netG)) to netG.load_state_dict(torch.load(opt.netG),strict=False) as #10 said to solve the following error, KeyError: 'unexpected key "tran_dense.dense_block1.denselayer1.norm.1.weight" in state_dict'

I succeeded running demo.py, but it seems that the outputs were not enhanced, as the following samples shows,

10_dcpcn

My environment: Ubuntu 16.04 Pytorch v0.3.1 with cuda 9.0 Python v3.5.2

BTW, would you like to remove the unused code. It really costs me some time to understand them. :smiley:

Thanks a lot!

hezhangsprinter commented 5 years ago

I have no idea what happens here since there is no useful information given to me.

But make sure you follow my code since I put some results obtained from my demo.py. For example, you should use net.train(). Please please go through my code in detail to make sure you are running properly.

xiaofeng94 commented 5 years ago

This problem is caused by my modification on netG.load_state_dict(torch.load(opt.netG)).

It seems that the author used torchvision v0.2.0 or earlier. However, the latest torchvision v0.2.1 has changed the the name 'norm.1' to 'norm0' in class _DenseLayer and thus "tran_dense.dense_block1.denselayer1.norm.1.weight" is not contained.

So, I change torchvision to v0.2.0 and everything go on well.

huoguangdiandian commented 5 years ago

I find a new idea to solve the problem. In fact, you do not change the version of torch. Only change the method that loading the model instead of original code.

CBD88 commented 4 years ago

I find a new idea to solve the problem. In fact, you do not change the version of torch. Only change the method that loading the model instead of original code.

Can you please share the code?