dataset
Closed this issue · 1 comments
miaolin968 commented
Is depth information used during training?
btsmart commented
Hello @miaolin968,
Yes, during training we use ground truth depth information and camera poses to supervise our reconstructions. If you have access to camera poses but not depth, you may wish to check out other methods such as PixelSplat and MVSplat, which introduce methods for handling the lack of depth.