Saliency map and deep learning based efficient facial emotion recognition technique for facial images

Detecting and recognizing facial emotions from human facial movements is one of the most important research area in human-computer interaction. In this paper, a saliency map and deep learning-based facial emotion recognition technique for facial images is proposed. The proposed model, contrast-limited adaptive histogram equalization (CLAHE), generative adversarial network (GAN), saliency map, and bilateral filter are used for data pre-processing. A deep convolutional neural network is trained using the Nadam optimizer to recognize facial emotion. After this, the proposed technique is tested on the JAFFE, CK+, and FER+ benchmark datasets. The maximum accuracies achieved with the proposed technique are 97.7%, 99.9%, and 84.2% for the JAFFE, CK+, and FER+ datasets, respectively.