Jacob Munkberg

I just created a new conda env, and installed PyTorch 1.11.0 in there, following the installation steps in the readme exactly. Everything runs fine on my end, and I don't...

Thanks @crud89 , Yes, we added the indexing parameter to remove the warning in PyTorch 1.10+

Hello, - More images should work out of the box. We pre-load datasets into CPU memory for perf (to avoid image parsing in the training loop), but also the preloading...

Yes, you need to adjust the camera frustum. We have a function `util.perspective_offcenter` that does this. We tested random cropping at some point but removed it for the public release....

Thanks @evenfh , We haven't tested with CO3D (yet), so I cannot quickly answer your question. Is the format of CO3D similar to the DTU MVS format? We did write...

Hello, After the first pass, we run [xatlas ](https://github.com/jpcy/xatlas) to create a UV parameterization on the triangle mesh. If the first pass failed to create a reasonable mesh, this step...

Hello @bonoshunki Yes, that looks worse than expected. With 189 images with good 3D coverage, I would expect higher quality, in line with the other results in the paper, such...

This seems to be an issue with tiny-cuda-nn, not nvdiffrec. I would recommend to report or check for solutions in the tiny-cuda-nn github repo: https://github.com/NVlabs/tiny-cuda-nn Also, as stated in our...

Hello, your `train_res [800, 800]` parameter must match the dataset image size, e.g., `DatasetNERF: 143 images with shape [3024, 4032]`, so I would recommend to either increase train_res in the...

@iraj465 Just to double check, are you using the exact same resolution on `train_res` and your image sizes? If you downscaled [3024,4032] to [756, 1008] then in the config, set...