search

ASolitaryMan / HFLEA

FRAME-LEVEL EMOTIONAL STATE ALIGNMENT METHOD FOR SPEECH EMOTION RECOGNITION
16 stars 0 forks source link

Problems with dataset division #1

Open Jiafei127 opened 6 months ago

Jiafei127 commented 6 months ago

For The format of training, validation, and test sources, is it done according to This means that the data of four sessions are used as training data; the data of one speaker from the remaining session is used as the validation set and the data of other one speaker forms the testing set.? Can the authors provide the official format file? I'm stuck at this step when reproducing.

ASolitaryMan commented 1 month ago

I am very sorry for my mistake. I double check the method for dataset division. I follow the leave one session out cross validation method https://github.com/ECNU-Cross-Innovation-Lab/ShiftSER. Four sessions are used as training data, and one session for testing set. I will mark errors in the readme.