Converted to TF2

[SiddhiVTripathi]

Great Work @SystemErrorWang . I was trying to do an ablation study on this research paper. I wasn't able to due to lack of resources( specifically multiple GPUs).

Here is the implementation of White-box-Cartoonization in Tensorflow V2. The dependencies for scikit-image and joblib remain the same. tf_slim is supposed to be installed on the platform. This can easily be done using pip command as "pip install tf_slim".

[ntbk1919]

Where is layers.Py used. It seems to be redundant

[ntbk1919]

@SiddhiVTripathi Could you tell if layers.py is being used anywhere at all

[SiddhiVTripathi]

Where is layers.Py used. It seems to be redundant

layer.py has functions like conv_spectral_norm that are used in the network. For example, conv_spectral_norm is used in the discriminator. See here line 101.

[ntbk1919]

Where is layers.Py used. It seems to be redundant

layer.py has functions like conv_spectral_norm that are used in the network. For example, conv_spectral_norm is used in the discriminator. See here line 101.

Thanks for the info and I'm grateful for it. Similarly,only disc_sn is being used,not disc_bn and disc_ln Also only unet_generator is being used ,not generator function.,right? Or have I made a mistake in code analysis?

[SiddhiVTripathi]

Where is layers.Py used. It seems to be redundant

layer.py has functions like conv_spectral_norm that are used in the network. For example, conv_spectral_norm is used in the discriminator. See here line 101.

Thanks for the info and I'm grateful for it. Similarly,only disc_sn is being used,not disc_bn and disc_ln Also only unet_generator is being used ,not generator function.,right? Or have I made a mistake in code analysis?

No, your code analysis is quite correct. I think one can use the U-Net generator in order to get similar results as authors have achieved in the paper, otherwise one can use the generator functions accordingly.