Matthew Tancik

Currently there is no option to upload record3d data in the colab. Depends on #854

For consistency, we should add an option in `ns-process-data` that takes in a record3d dataset. We can then remove the record3d dataparser and only use the nerfstudio dataset format.

Print a summary warning to users if colmap does not generate many matches.

**Describe the bug** Clicking the "eye" in the camera list moves the green camera, but does not change the time step. Therefore when interacting with the scene it snaps back...

Currently if you have the viewer enabled, the model will not quit when it is done training. We should add a command line argument that enables this behavior.

When you enable looping, the camera path gets pinched at the start/end  The behavior in an older version is desired 

Adding support for polycam captures. To export data, Polycam must be in developer mode.

Trying to run `ns-train` with an absolute path fails to run.

It would be nice to be able to choose the interpolation method when setting the background image. Ideally a interface like, ``` set_background_image(image: numpy.ndarray, format: Literal[png, jpeg] = 'jpeg', jpeg_quality:...

Seems supported in Leva: https://leva.pmnd.rs/?path=/story/inputs-image--image