Project aimed to compare the performances of MLP, SVM and CNN algorithms in detecting the facial emotions. It also uses a pre-trained model and evaluates its performance in facial emotion recognition. The highest accuracy was achieved by the pretrained CNN model in detecting emotions.
Dataset - RAF Dataset
File Contents -
- SVM_MLP_Train.ipynb - Google Colaboratory file to train MLP and SVM models. Both models are tried with different combinations of feature descriptors like SIFT, and HOG.
- CNN_Train.ipynb - Google Colaboratory file to construct and train a Convolutional Neural Network.
- Test_Emotion_Recognition_RAFDataset_SVM_MLP.ipynb - Google Colaboratory file to test the MLP, SVM and CNN models on the test RAF dataset and note down the accuracies of the models.
- Test_Emotion_Recognition_Video_CNN.ipynb - Google Colaboratory file to test the MLP, SVM and CNN models on a YouTube video.