Some question of evaluating OccNerf on OcMotion datasets.

[imabackstabber]

Hi, it might sound like a silly question... But I wonder whether the PSNR_full and SSIM_full reported in the original paper are somehow unuseful...?

Since we only focus on the foreground. Apply PSNR function on the whole image is somehow misleading, since the human body on the gt_img is masked out, and rendered results are, in fact, full body.

I found this in another issue

So to my understanding, when evaluating on OcMotion, we just need to, or to say, have to compute only PSNR_vis and SSIM_vis?

I fully understand if there is some issue going on with the OcMotion table since one can't change certain content in the camera-ready version of their paper... Please let me know and it will help a lot!