search

taesungp / contrastive-unpaired-translation

Contrastive unpaired image-to-image translation, faster and lighter training than cyclegan (ECCV 2020, in PyTorch)
https://taesung.me/ContrastiveUnpairedTranslation/
Other
2.23k stars 417 forks source link

Using custom dataset #112

Open oliver0922 opened 2 years ago

oliver0922 commented 2 years ago

Thanks for great repository!

I want to use CUT model to change day-time datasets to night-time datasets using bdd1000k images.

Is it possible to change using CUT model?

taesungp commented 2 years ago

Yes definitely.

  1. Please prepare the datasets in a directory. The directory should contain 4 subdirectories trainA trainB testA testB.
  2. Start training using python train.py --dataroot [path-to-the-dataset] --name bdd1000k_CUT --CUT_mode CUT
oliver0922 commented 2 years ago

Thanks for your reply. By the way I want to change day time to night time using FastCUT. After training the model using bdd1000k, I want to apply the model on KITTI datasets. As you know, KITTI datasets are obtained only at day-time, so I dont have testB folder. Do you have any suggestions? or Do I have to use SinCUT?