search

NVlabs / neuralangelo

Official implementation of "Neuralangelo: High-Fidelity Neural Surface Reconstruction" (CVPR 2023)
https://research.nvidia.com/labs/dir/neuralangelo/
Other
4.4k stars 389 forks source link

Does neuralangelo is able to deal with the datasets captured by multiple cameras? #120

Open ZiyangYan opened 1 year ago

ZiyangYan commented 1 year ago

Hi, Thanks for your excellent work. Now I have a dataset which is captured by 5 cameras in the air. So some of the images have different resolution. Does neuralangelo can deal with the datasets captured by multiple cameras?

mli0603 commented 1 year ago

Hi @ZiyangYan

Thank you for your interest in the project! The current dataloader only supports single camera. However, it is possible to extend this to multiple cameras. If you look at the transforms.json file, a single camera intrinsics is assumed. You can instead move the intrinsics info under each camera for more multi-camera support. Feel free to submit a PR if you have something working :)

FrankSpalteholz commented 1 year ago

Also, big kudos for this game-changer and on top to make this public! I'm also interested if I could provide image-data by using 7 GoPro's with the exact same image-size-settings? I know that technically there are also differences in the intrinsics, but can you say something if those (even if small) would make the quality reduce by a lot? Have you tested something similar as setup?

mli0603 commented 1 year ago

Hi @FrankSpalteholz

I would extend the current implementation to support multi-cam instead of assuming that they all have the same intrinsics :) I haven't tested something like yours, but I am not sure if this is safe to assume.