Emotion Detection using Facial Expressions

Paper

Submitted to FTNCT 2021. Cite as: "to be included"

Datasets

CK+

Emotion	#Images
Angry	45
Contempt	18
Disgust	59
Fear	25
Happy	69
Sadness	28
Surprise	83

P. Lucey, J. F. Cohn, T. Kanade, J. Saragih, Z. Ambadar and I. Matthews. "The Extended Cohn-Kanade Dataset (CK+): A complete dataset for action unit and emotion-specified expression," 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition - Workshops, 2010, pp. 94-101. doi: 10.1109/CVPRW.2010.5543262.

JAFFE

Emotion	#Images
Angry	30
Disgust	29
Fear	32
Happy	30
Neutral	31
Sadness	30
Surprise	30

Michael J. Lyons, Shigeru Akamatsu, Miyuki Kamachi, Jiro Gyoba. Coding Facial Expressions with Gabor Wavelets, 3rd IEEE International Conference on Automatic Face and Gesture Recognition, pp. 200-205 (1998). http://doi.org/10.1109/AFGR.1998.670949 Open access content available at: https://zenodo.org/record/3430156

Proposed Methods

PCA+SVM

The models for CK+ and JAFFE datasets are available in pca_svm_model.ipynb.

JAFFE

For JAFFE we used the first 130 principal components.

CK+

For CK+ we used the first 120 principal components.

CNN

The same model was used on both the datasets. The model and prediction for CK+ and JAFFE datasets can be found in fer_cnn_ck+.ipynb and fer_cnn_jaffe.ipynb respectively.

Resuts

Method	Dataset	Accuracy(%)
PCA+SVM	JAFFE	95.35
PCA+SVM	CK+	81.81
CNN	JAFFE	87.50
CNN	CK+	83.84