image-generation

Hi! I was wondering if it is possible to read data from AWS S3 to train a model such as BasicVSR.

I have been reading the docs and all indicates it could be done by modifying my config file, and setting the backend option to "ceph" instead of the default "disk".

Then I sett the paths to:
lq_folder = "S3://my-bucket/my-dataset/lq_folder"
gt_folder = "S3://my-bucket/my-dataset/gt_folder"

But I ca

The quality of images is higher when the number of pixels from the source and target match 1:1. When a larger source is provided, it could be randomly cropped if a command-line argument is set (likely on by default).