sugar2019 / facial-expression-recognition-in-wild-datasets

use softmax, weighted softmax, weighted cluster,weighted Island loss etal on RAF-DB and FERPLUS2013,based on Pytorch

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

facial-expression-recognition-in-wild-datasets

use softmax, weighted softmax, weighted cluster,weighted Island loss etal on RAF-DB and FERPLUS2013,based on Pytorch

run train_RAF_sm_pre.py and train_RAF_Wsm_pre.py will save trained models, you can preserve the last file in every model_file, which has the best test accuracy. Then you can run test_cm.py to produce confusion matrix on RAF-DB. Don't forget modify the path of model and save img. If you want to see the confusion matrix on FERPLUS,then you run test_generaltive_onFERPLUS.py

The other train_XXXX.py all use "centers",so you run them to produce correlate model. Then you can run test_cm_center.py to produce confusion matrix on RAF-DB. Don't forget modify the path of model and save img. If you want to see the confusion matrix on FERPLUS,then you run test_generaltive_onFERPLUS_center.py

About

use softmax, weighted softmax, weighted cluster,weighted Island loss etal on RAF-DB and FERPLUS2013,based on Pytorch


Languages

Language:Python 99.8%Language:Shell 0.2%