Having problems in understanding the implement of 10-fold Cross validation ?

WuJie1010 / Facial-Expression-Recognition.Pytorch

A CNN based pytorch implementation on facial expression recognition (FER2013 and CK+), achieving 73.112% (state-of-the-art) in FER2013 and 94.64% in CK+ dataset

MIT License

1.79k stars 550 forks source link

Having problems in understanding the implement of 10-fold Cross validation ? #16

Closed juneweng closed 5 years ago

juneweng commented 5 years ago

Hi, WuJie！this is a great work! I am interested in tuning the model.

I am having problems while debugging "mainpro_CK+.py". parser.add_argument('--fold', default=1, type=int, help='k fold number')

My questions:

fold default is 1,while I has not input,It turn to 10.I really want to know how to realize.
Could you please give me more detail about the implement of 10-fold Cross validation?

Thank you in advance!

WuJie1010 commented 5 years ago

please use "python k_fold_train.py"

juneweng commented 5 years ago

Thank you for your prompt!

there are my understands:for 1~9 fold , we count test number in order,for 10 fold,we count in the reverse order,In that way,we ignore some overlaps or discard some pictures,such as angry(135),we count 120,disgust (177) we count 180.But It doesn't matter.every time,we choose the fix test number,and the rest for train.Do I understand correctly?

Recently, I am stuck in data augment in CK+, your work help me a lot.Thank you once again.

WuJie1010 commented 5 years ago

Yes

juneweng commented 5 years ago

Thank you！