search

qianlim / POP

Official implementation of the ICCV 2021 paper: "The Power of Points for Modeling Humans in Clothing".
https://qianlim.github.io/POP
Other
187 stars 20 forks source link

chamfer distance #13

Closed caiyongqi closed 2 years ago

caiyongqi commented 2 years ago

Hi. You don't seem to be using the chamfer distance directly as the loss, would the performance be better? https://github.com/qianlim/POP/blob/a05404dd44173809abfee4e2ebef57bbc524cfde/lib/train.py#L87 https://github.com/qianlim/POP/blob/a05404dd44173809abfee4e2ebef57bbc524cfde/lib/train.py#L98

qianlim commented 2 years ago

I'd say the performance difference is very small when the points are dense. It's computed this way due to some historical reasons. Just using the plain chamfer distance should also be fine.

caiyongqi commented 2 years ago

I'd say the performance difference is very small when the points are dense. It's computed this way due to some historical reasons. Just using the plain chamfer distance should also be fine.

Thank you! Now I close this issue.