We present a convolutional neural network (CNN) for 2D + 3D featurebased facial expression recognition approach and present its performance using BU-3DFE database. Our network consists of two CNNs: one for the 3D face shape and the other for the face appearance with color in order to achieve efficiency and robustness. The network consists of three convolutional layers including max pooling as well as normalization layers, and two fully connected layers, totaling over 26.5 million parameters and 45504 neurons. A 6-way softmax is used for the outputs on the final layer. Performance evaluation suggests that the facial expression recognition rate reaches to excellent 92%.
CITATION STYLE
Huynh, X. P., Tran, T. D., & Kim, Y. G. (2016). Convolutional neural network models for facial expression recognition using BU-3DFE database. In Lecture Notes in Electrical Engineering (Vol. 376, pp. 441–450). Springer Verlag. https://doi.org/10.1007/978-981-10-0557-2_44
Mendeley helps you to discover research relevant for your work.