Can you modify for train_1, train_2 ... like val_1, val_2 ...?

[ziippy]

I want to train with separated folders like train_1, train_2.

I checked the .yaml support multiple validation directory like val_x. but, train_x is not supported. Is there any plan for supporting that?

my GT is divided separated with each domain. (load, bird, car, ... and so on) So, If you can. please do that.

[chxy95]

This repo is based on BasicSR that supports this function. You can put multiple folders in the same directory and use this directory as the dataset path. Then 1) you can rewrite a data loader file by setting the recursive to True for the scandir function as scandir(path, recursive=True, full_path=True) https://github.com/XPixelGroup/HAT/blob/4a77c403a0982fbb141453680c4c07569f91447b/hat/data/imagenet_paired_dataset.py#L35 2) or you can generate a meta info file according to your needs by using generate_meta_info.py to scan the data.