in this project i was about studying classical machine learning models by creating simple classifier between 7 human emotions (neutral, anger, contempt, disgust, fear, happiness, surprise). main goal of the algorithm is to predict the emotions on a person's face photograph using some labeled dataset.
in this case i'm taking dataset containing human faces photographs where they were asked to show 7 emotions -> preparing it for machine learning methods (cropping, augmentation, histogram equalization) -> extracting gabor features -> applying PCA+LDA combo -> using some basic ml models (dt, rf, adaboost, gradient boosting, svm, knn and naive bayes)
folder with images you can download here
here is labeled dataset as .xlsx file as a label - filename table
the only problem about dataset used is disbalance in quantity of photoghaphs in different emotions samples (there are much more neutral examples) and that is why augmentation is requiered