Closed MCLYang closed 4 years ago
The training step for all models is here: https://github.com/erikwijmans/Pointnet2_PyTorch/blob/master/pointnet2/models/pointnet2_ssg_cls.py#L125
I am using F.cross_entropy
(which is softmax + nll_loss).
thank you
Hi @erikwijmans, I am so new to pytorch-lighting. I did not find the loss function from the code of trainer. What is the loss function for the semantic segmentation? From other implementation for pointnet++, I found its just like
F.nll_loss()
but I still want to confirm if your version is usingF.nll_loss()
or you add the regularizer?