search

amineHorseman / facial-expression-recognition-using-cnn

Deep facial expressions recognition using Opencv and Tensorflow. Recognizing facial expressions from images or camera stream
GNU General Public License v3.0
461 stars 142 forks source link

about the accuracy #14

Closed dagongji10 closed 5 years ago

dagongji10 commented 5 years ago

Hi, I have some problems in optimizer param. I run the optimize_hyperparams.py with max_evals=20, but still cannot get the acc of 75.1%. In fact, my results always under 70%. Is the value of max_evals setting too small? I use 5 emotions(angry, happy, sad, suprise, neytral). Could you help me?

amineHorseman commented 5 years ago

Well, that's pretty strange since I got everytime a score over 70% for the model B, did you change some parameter? can you print your parameters.py file?

dagongji10 commented 5 years ago

Well, that's pretty strange since I got everytime a score over 70% for the model B, did you change some parameter? can you print your parameters.py file?

In fact， I changed the database and network. The labels of Fer2013 are not exactly and size of image is too small, so I use RAF-DB to replace Fer2013. Finally, I can get 83~84% acc using 7 emotion.