2DGS Evaluation with scannet++ iPhone dataset

[Rickyeeeeee]

Thank you for your excellent work and contributions!

I am particularly interested in the evaluation process described in your paper. From my understanding, nerfstudio performs image undistortion each time the camera extrinsics and intrinsics are loaded. Given that the scannet++ iPhone datasets use the OPENCV camera model, this undistortion step is conducted during both training and evaluation.

Could you share how you conducted experiments on 2DGS, which only supports the PINHOLE camera model? Did you pre-save undistorted images from nerfstudio and use them in your evaluations?

Thank you for your time and insights.