Giters

yl-1993 / learn-to-cluster

Learning to Cluster Faces (CVPR 2019, CVPR 2020)

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

The way to get 256-dimension features

szh-bash opened this issue · comments

commented

Hello,
I'm wondering whether you got 256-dimension features from Higher-Dimension features with dimensionality reduction algorithm or Network Output?
If you were using dimensionality reduction algorithm, what's the name of it.
Or something else?

Thank You :)

commented

@szh-bash Hi, we obtain the 256-dimension features with ResNet-50 used for face recognition, which adds a bottleneck fc layer with 256 outputs between backbone and classifier. The detailed structure of our network can be found in hfsoftmax.