ethanweber
commented
1 year ago Hi @h8c2, thanks for noting this. Do you imagine that this function call should look at termination_depth and mask out regions where GT depths == 0 in the loss?
We might not get to this soon but are open to pull requests. I think it's a good addition as long as the documentation makes clear what is happening.
aayushg55
I am using nerfstudio working on Lidar-Image Dataset. I convert the lidar points to depth images and follow the guide https://github.com/nerfstudio-project/nerfstudio/blob/main/docs/quickstart/data_conventions.md to train my model. However, It seems like the code doesn't apply a mask to the zeros area of the depth images. This would be very helpful if you provide support related to sparse depth training.