thanks for releasing the code, I really appreciate your work!
I want to try your code on the ScanNet dataset and get above error message after implementing a ScanNet dataloader.
Couuld you please answer the following? :)
ScanNet trajectories are not 360 / outside-in trajectories like in Tanks&Temples dataset.
Hi,
thanks for releasing the code, I really appreciate your work! I want to try your code on the ScanNet dataset and get above error message after implementing a ScanNet dataloader. Couuld you please answer the following? :)
ScanNet trajectories are not 360 / outside-in trajectories like in Tanks&Temples dataset.
Do you think a simple rescaling of the poses to be within [-1,1] will be enough to make your method run on such poses? By this I mean multiplying each pose's position with e.g. 0.25 to make them closer together (similar to the procedure in instant-ngp: https://github.com/NVlabs/instant-ngp/blob/master/docs/nerf_dataset_tips.md). See here for the current implementation: https://github.com/lukasHoel/Stylizing-3D-Scene/blob/fc203e0286e33d684f68f4f2d2000a1db827a143/dataset/dataloader.py#L154
Or will it be more difficult and we need to add background models, a larger ray sampling step size, etc. ?
Thanks in advance! Lukas