How to test on real-world images about cam2bev

[lreiher]

You are correct that our proposed methodology takes semantically segmented camera images as input. As stated in our paper, we have used an in-house semantic segmentation model for the real-world examples shown in the paper.

We don't specifically recommend any model for semantic segmentation, but naturally, a better semantic segmentation model would also provide better input for Cam2BEV. There should be plenty of semantic segmentation methods publicly available, one starting point might be to take a look at paperswithcode.