Joker316701882 / Additive-Margin-Softmax

This is the implementation of paper <Additive Margin Softmax for Face Verification>
492 stars 149 forks source link

the face images is 160*160,train result will be good? #11

Closed deeplearninggogogo closed 6 years ago

deeplearninggogogo commented 6 years ago

why resize images is 112*96? thanks

Joker316701882 commented 6 years ago

@huyazhou-2018 Historical reason. It should be good with 160 but not for sure, because of the valid receptive field concern, maybe you need to use deeper network.