search

neka-nat / image_completion_tf2

"Globally and Locally Consistent Image Completion" with Tensorflow2 Keras
31 stars 11 forks source link

where did you use the joint loss? #1

Closed micklexqg closed 6 years ago

micklexqg commented 6 years ago

thanks for your work. I am running your code and have some questions. where did you use the joint loss in your code? And I notice the weight for discrimitor loss is 0.0004 in the original paper. where did you set the parameter? thanks, any help appreciated!

neka-nat commented 6 years ago

Thanks @micklexqg . My code is referring to glcic and their implementation seems to have a similar issue. I made the new branch implemented according to the paper, so please check it. https://github.com/neka-nat/image_completion_keras/pull/2

micklexqg commented 6 years ago

thanks for quich reply. i checked it, but maybe there is still no update for joint loss...

neka-nat commented 6 years ago

I updated again. https://github.com/neka-nat/image_completion_keras/pull/2/files#diff-e44f4a60e89e820dde9bb27afa634965R135

micklexqg commented 6 years ago

@neka-nat , thank you, need i copy the code from the given url?(download from the github clone is still the old version ). another question, is there any pretrained model, I think uploading a pretrained model is better.

neka-nat commented 6 years ago

If you use new version branch, please run the follow commands.

git clone https://github.com/neka-nat/image_completion_keras.git
cd image_completion_keras
git reset origin/separate_c_d --hard

Sorry, I can not prepare the pre-train model yet. Please wait several weeks.

micklexqg commented 6 years ago

@neka-nat ,thanks, if it is finished ,please inform us. thanks again!