Test on Ev-REDS

[renliao]

Hello, thank you for your contribution, I want to quantitative comparisons under different spatial scales on Ev-REDS, so I train with roi-size=64 on evreds_train.yaml, and I test with roi-size=64 on evreds_test.yaml, does it means R=4? if I want to test with R=1, should I test with roi-size=256 on evreds_test.yaml? Thank you, I wish you have a happy life.

[XiangZ-0]

Hi, thanks for the question. roi-size is a parameter that only affects training, so you don't need to care about it during inference. When testing Ev-REDS, the default setting in evreds_test.yaml will yield the results of R=4. To compute the results of R=1, you can downscale the images in npy files, i.e., blur1, blur2, sharp_imgs, to the same spatial scale of events. Afterward, modifying scale_factor=1 in evreds_test.yaml. Then it should work.