Coordinate system used by TensoRF as opposed to that of NeRF and instant-ngp

[NagabhushanSN95]

trafficstars

Hi,

In issue-46, you mention that tensorf uses a different coordinate system. Can you please elaborate on that? From the code, I understand that NeRF uses the standard (x,y,z) system, but TensoRF uses the OpenCV (x,-y,-z) system. Is that all or is there some more difference?

Also, can I use TensoRF in the standard (x,y,z) system? To do this, will it work if I just need to transform the camera extrinsics to (x,y,z) system and multiply the 2nd and 3rd dimensions of directions by -1?