search

taylover-pei / SSDG-CVPR2020

Single-Side Domain Generalization for Face Anti-Spoofing, CVPR2020
224 stars 35 forks source link

Hyperparameters in config are the ones achieved results in the paper? #11

Closed XingyuJinTI closed 4 years ago

XingyuJinTI commented 4 years ago

Hi there, I achieved different results with your current config files and three of them are lower performed.

Screenshot from 2020-09-01 23-03-40

Mine was like this Screenshot from 2020-09-01 23-03-51

I just wonder if Hyperparameters in config are the ones achieved results in the paper? Many thanks

1relia commented 4 years ago

我也遇到这了这个问题 image 请问你是默认参数运行的么

taylover-pei commented 4 years ago

Sorry for answering late.

The hyperparameters are the same as our original paper.

Two points need to be confirmed:

  1. Did you follow the code prompts and use the specified MTCNN code for face detection and alignment?

  2. Is the frames of each video selected for training are the same with us?

If it is due to data processing or experimental environment problems, I suggest you to tune some parameters by yourself to find a set of optimal parameters suitable for your data set and environment.