Dominik Werman
Dominik Werman
Have you found any solution or source of error? Do you use work with the preprocessed images (D2-Net) for the megadepth dataset?
@xushangnjlh Hey, I am also observing this behavior. Where you able to spot an error in you approach? My scannet evaluation also does not reach the exspected values.
Sorry, I misunderstood the size warning for the megadepth dataset. It seems to be the correct dataset. But still the issue of a high loss of 4.7 remains after all...
I trained the network on the coco dataset only, which did not result in an loss improvement.
@dinglei0719 Hey, I guess the loss definition might be a causing a relative high total loss. But besides the loss feeling pretty high the network does perform as expected. I...