Question about Scannet++ preprocessing

[saulgooodman]

Hello, I would like to test the capabilities of SplatAM on ScanNet++, but I find the preprocessing steps for ScanNet++ quite confusing. My understanding of the overall process is as follows:

1. Render a distorted depth map based on the mesh and poses obtained from the GT mesh.
2. Undistort the RGB and depth maps to obtain undistorted RGB and depth maps.
3. Use the undistorted RGB and depth maps as input data for training.

However, I'm not sure how to perform the undistortion. I noticed that https://github.com/Nik-V9/scannetpp mentions two methods for undistortion, and the results of these methods are different. How should I choose the undistortion method for RGB and depth? Thank you!