Custom data

[katted8]

Hello,

Thank you for your amazing work and for sharing the code!

Suppose I already have paired RGB and depth images of each frame of a multi-view video. How should I convert this data into the structure required by your method? Specifically, I'm a bit unclear about the structure and purpose of the files in the masks_humanrf and sparse folders, as well as the structure of the rgb_cameras.npz file.

Thanks for your help!

[chengwei-zheng]

Hi, sorry for the late response. You can simply go through the data processing on the ActorsHQ dataset, and you will find everything there.

The masks_humanrf folder contains masks for the captured objects, and the sparse folder contains the camera parameters in COLMAP format. For the usage of rgb_cameras.npz, you can refer to the following script:
https://github.com/eth-ait/GauSTAR/blob/master/data_process/render_depth_from_mesh.py#L27