SwinIR config settings for GAN training

[PeterVennerstrom]

The GAN configs for SwinIR Real World SR appear to only change the experiment name after inheriting the psnr settings.

_base_ = ['swinir_psnr-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py']

experiment_name = 'swinir_gan-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py'
work_dir = f'./work_dirs/{experiment_name}'
save_dir = './work_dirs/'

I haven't found any settings related to the GAN model definition or loss. The setting to load the psnr trained model for fine-tuning with GAN loss is also not seen in the child config.

Any help locating these settings would be appreciated!

[Z-Fran]

We only support the inference of swinir and inference only needs generator of GAN, so two configs are the same. If you want to train the model, you can use the official codebase of swinir from https://github.com/JingyunLiang/SwinIR/tree/main and https://github.com/cszn/KAIR/tree/master .

[PeterVennerstrom]

Thank you for your quick response.