Documentation for camera coordinate frames and creating the json

[akshay-krishnan]

Is your feature request related to a problem? Please describe. What is the camera coordinate frame used by nerfstudio? Colmap uses +x -> right, +y -> down, +z -> forward. The code at this line makes me think that nerfstudio uses +x -> right, +y -> up, +z -> behind camera, like blender?

I would like to use camera poses from a different pipeline (synthetic images). I could create the json file that you create after pre-processing the colmap output, but I am confused what the coordinate system is. Specifically the lines 439-440 here show that the X and Y axes are being swapped, and z is flipped (again) ?

Some explanation of the camera coordinate frame would be very useful! Please point me to it if it exists already.

[ethanweber]

Hey @akshay-krishnan, you are correct in your interpretation. I'm adding the docs in PR #889.

[akshay-krishnan]

Thank you @ethanweber !