Daniel Crispell

add support for per-image mask filenames in transforms.json

2

Add support for per-image mask filenames via a `"mask_path"` key in each element of `"frames"` in `transforms.json`. The `"mask_path"` key must be present for all frames or zero frames.

This PR bring the behavior of [collate_image_dataset_batch_list](https://github.com/nerfstudio-project/nerfstudio/blob/32939e0d4198e23e9cef2fb64718cc1d263cab61/nerfstudio/data/pixel_samplers.py#L218) in line with that of [collate_image_dataset_batch](https://github.com/nerfstudio-project/nerfstudio/blob/32939e0d4198e23e9cef2fb64718cc1d263cab61/nerfstudio/data/pixel_samplers.py#L163C9-L163C36) by sampling pixels from all data in the batch, not just "image".

…point I ran into some problems using the output vtkXMLImageData file - The values are correct, but the locations don't match the original mesh vertices. This seems to fix the...

bugfix: handle multiple camera types in a single batch

2

Training with mixed batches of camera types seems to be broken. The camera types are looped over, but the same type will trigger the if/else each time through the loop...