search

nileshkulkarni / csm

Code release for "Canonical Surface Mapping via Geometric Cycle Consistency"
https://nileshkulkarni.github.io/csm/
186 stars 31 forks source link

Quaternion Azimuth-Elevation predictor #20

Closed tejank10 closed 4 years ago

tejank10 commented 4 years ago

Hello Nilesh,

In the quaternion azimuth-elevation predictor, linear layer output is never used. Is that intended? On this line.

Thanks.

nileshkulkarni commented 4 years ago

Hi @tejank10, Thanks for pointing that out, you could have the model work both ways. You could have a linear layer or not. It probably helps (at least won't harm to have the linear layer). It was not intended to be like this, but all our models are trained without using it. There are multiple linear layers here so it would be equivalent to adding another linear layer.

Best, Nilesh