About Fusion of Image and LiDAR instead of only-LiDAR

[fengjiang5]

Hello, I am interested in your work on 2DPASS, which is a method for semantic segmentation on LiDAR point clouds with 2D priors. I wonder if you have tested adding an image branch as a fusion method on nuscenes, instead of using only point clouds.
Could you please share your insights or results on this? Thank you very much.