Hello,
Thank you for open-sourcing this work. I have a few questions regarding it. When I tried to reproduce this work on a similar snowykitti dataset, I found that I was able to train the model normally, but when using the trained weights for evaluation, all predictions were incorrect, with both TP and FN being 0. Interestingly, when I used a model pre-trained with WADS without changing anything else, I was able to get reasonable results, although the recall also tended towards 0.