search

yuval-alaluf / SAM

Official Implementation for "Only a Matter of Style: Age Transformation Using a Style-Based Regression Model" (SIGGRAPH 2021) https://arxiv.org/abs/2102.02754
https://yuval-alaluf.github.io/SAM/
MIT License
632 stars 151 forks source link

The result when using encoder_type "BackboneEncoderUsingLastLayerIntoWPlus" is almost identical with the result when using "GradualStyleEncoder" #30

Closed datbu178 closed 3 years ago

datbu178 commented 3 years ago

Hi. I tried training using encoder_type "BackboneEncoderUsingLastLayerIntoWPlus". The other parameters are kept the same with the old settings. As a result, the loss function on the validation set is almost equal. And testing on some real images also gives almost identical results. So is it necessary to use encoder_type "GradualStyleEncoder"? When it requires significantly more computation. Thank you!

yuval-alaluf commented 3 years ago

Hi @datbu178, You are correct that you may be able to achieve similar results with a simpler backbone for the encoder than the FPN-based encoder.