search

jonbarron / robust_loss_pytorch

A pytorch port of google-research/google-research/robust_loss/
Apache License 2.0
656 stars 88 forks source link

is adaptive loss suitable for 3d shape reconstruction? #30

Closed qiminchen closed 2 years ago

qiminchen commented 2 years ago

Hi, I'm wondering if the adaptive loss is suitable for 3d shape reconstruction, especially for implicit representation, e.g, predicting inside/outside values? Thanks

jonbarron commented 2 years ago

Hard to say, if it's a regression task where you don't know the noise model a-priori, it could be helpful. If you have a noise model then this probably isn't a good idea.