Did L_var term in discriminative loss forget to divide the batch_size?

pqhieu / jsis3d

[CVPR'19] JSIS3D: Joint Semantic-Instance Segmentation of 3D Point Clouds

https://pqhieu.com/research/cvpr19/

MIT License

176 stars 36 forks source link

Did L_var term in discriminative loss forget to divide the batch_size? #19

Closed kyuusaku closed 4 years ago

kyuusaku commented 4 years ago

https://github.com/pqhieu/jsis3d/blob/d4c27e6492e70c0e9c07e04c0bc348348892c674/losses/discriminative.py#L61

      for i in range(batch_size):
            n = size[i]
            loss += torch.sum(var[i,:,:n]) / torch.sum(masks[i,:,:n])
      loss /= batch_size

pqhieu commented 4 years ago

Hi @kyuusaku,

Good catch, let me correct the code. I'm not sure it will have any impact on the performance though.