train on custom dataset about pasd HOT 4 OPEN

Hi, thanks for this excellent work! I would like to try on my own dataset (like hr_512, lr_128), could you please tell me which dataloader I should use to change to my own dataset path?

Thanks

If you are using a paired dataset, you can use localdataset and modify the dataloader in the training code. If you are using webdataset, you need to convert the images to a unified format such as JPG and then compress them into a TAR file. This is because webdataset is a library that reads TAR files. Once you specify the path correctly, it should work successfully.

from pasd.

thanks for the quick reply. I am using paired dataset (one hq image and corresponding one lq image). I notice that in localdataset, there is lq_folder which needs to be changed to my own dataset path, could you please tell me where I can set the hq_folder?

from pasd.

thanks for the quick reply. I am using paired dataset (one hq image and corresponding one lq image). I notice that in localdataset, there is lq_folder which needs to be changed to my own dataset path, could you please tell me where I can set the hq_folder?

I am not a contributor to the code, but I believe that here you need to construct hq using the same method and pass it into the example object. You can refer to the writing style of webdataset to ensure that the same results are returned.

from pasd.

FYI, I have updated the dataloader/localdataset.py.

from pasd.