search

roimehrez / contextualLoss

The Contextual Loss
http://cgm.technion.ac.il/Computer-Graphics-Multimedia/Software/Contextual/
489 stars 78 forks source link

sample test issue #5

Closed kwkwvenusgod closed 5 years ago

kwkwvenusgod commented 6 years ago

I am trying to use the pretrained model for quick start. But it is producing some results which I don't think are correct. 00010_out As far as I understand, the provided pre trained model is to warp trump animation to some target images. So it should be like the one in Fig 1 (left) in https://arxiv.org/pdf/1803.02077.pdf

I really appreciate you that you can give me some hints if there is something wrong I've done.

roimehrez commented 6 years ago

It is very hard to know what is the problem, you didn't give any details. It might be some version issue? I test it on Windows with tensorflow 1.8 or 1.3. scipy 1.1.0 pyton 3.5 pillow 5.1. please give more details

kwkwvenusgod commented 6 years ago

I tested using MACOS and linux both and both produced result like this.

For the config for Linux: python 3.5 tensorflow-gpu-1.8 scipy 1.1 pillow 5.1

For MACOS, the configuration is similar except that the tensorflow version is for cpu and python is 3.6.

I will try to use your own dataset in my machine and see.

roimehrez commented 6 years ago

did you solve it?

kwkwvenusgod commented 6 years ago

I retrain the model using your sample materials. The test result becomes fine using the retrained one. I think that should be the OS issue.

zhuangxinfa commented 5 years ago

I am trying to use the pretrained model for quick start. But it is producing some results which I don't think are correct. 00010_out As far as I understand, the provided pre trained model is to warp trump animation to some target images. So it should be like the one in Fig 1 (left) in https://arxiv.org/pdf/1803.02077.pdf

I really appreciate you that you can give me some hints if there is something wrong I've done.

move the checkpoints in the folder "single_im_CXt32_42_1.0_CXs42_1.0" the problem will be sloved