search

rcmalli / keras-vggface

VGGFace implementation with Keras Framework
MIT License
928 stars 417 forks source link

512 Features instead of 4096 #37

Closed ghost closed 5 years ago

ghost commented 6 years ago

Hi there, thank you for posting this implementation. I have a question. I am confused at why I seem to be getting 512 dimensional representations when I set include_top = False instead of 4096 dimensions. I can get 4096 dimensional representations by setting include_top = True and using the output from layer fc7, however I am curious about what is going on in the former case. The 512 dimension representations seem to perform better for the purpose of comparing face similarity too.

rcmalli commented 5 years ago

When you set the include_top=False the model does not have the fully connected layers. That is why you are only able to get 512 dimensions.