How can I train my own dataset?

[coallar]

Hi, thanks for your great job provided by DaTaoBao tech. !

what I want to know about is that how can I train my private dataset. It seems that gt image should be provided in this project？While, only dataset without gt info do I have, So, how can I solve it?

thanks!

[ImCharlesY]

Hi @coallar,

So sorry for the late reply!

Our project focuses on supervised learning where the paired reference image (or GT image) is required to provide the supervision. When dealing with datasets without reference images, some unsupervised learning strategies should be adopted, such as using the GAN (Generative Adversarial Networks) loss based on unpaired reference images. Unfortunately, this is out of the scope of our work, and we do not have a plan for it. Our related work TPAMI Ada-3DLUT and the MMEditing toolbox provide some support for adversarial training. You can refer to Ada-3DLUT/UnpairedTraining, MMEditing/GenerationUnpairedDataset, and MMEditng/CycleGAN for more information.

I'm going to close this issue. If you have any further questions, please feel free to comment.