Thanks a lot for the amazing work!
Regarding the ground truth albedo map, it seems that the SfSNet code to which you referred only provides output of size 128x128, whereas your network input is 512x512...
So I wonder, did you retrain SfSNet on larger image, or you just simply upsample the prediction from off-the-shelf SfSNet?
Hello Xiaoxing,
Thanks a lot for the amazing work! Regarding the ground truth albedo map, it seems that the SfSNet code to which you referred only provides output of size 128x128, whereas your network input is 512x512... So I wonder, did you retrain SfSNet on larger image, or you just simply upsample the prediction from off-the-shelf SfSNet?
Many thanks, Shiyang