iamhankai / ghostnet.pytorch

[CVPR2020] GhostNet: More Features from Cheap Operations
https://arxiv.org/abs/1911.11907
522 stars 116 forks source link

About SELayer #2

Closed WangDeyu closed 4 years ago

WangDeyu commented 4 years ago

After excitation layer, there is "clamp operation" but not "sigmoid", so why does it have this modification?

iamhankai commented 4 years ago

because clamp is faster than sigmoid.

WangDeyu commented 4 years ago

Well, does this modification affect performance?

iamhankai commented 4 years ago

The performance has no significant difference.

WangDeyu commented 4 years ago

OK, thanks!