Visual Emotion Recognition based on transfer learning technique using VGG16

Visual emotion recognition is one of the active topics nowadays. Recognizing emotions from a sequence of moving images still shows some difficulty in correctly detecting the exact features due to facial movement in the first place. Especially the movement of the mouth when pronouncing the sentence while producing emotions, which mainly affects the appearance of facial features. Thus, in this work, we focus on emotion recognition from facial expressions expressing speech. The deep neural network used in this work is VGG16 which is considered to be an effective neural network for detection and classification tasks, and can mainly be adaptable with transfer learning, technique. The presented method is conducted on the Video-speech category where we work on the detection of six classes of emotions which are: neutral, calm, happy, sad, angry and fearful, where the precision obtained is 78.12%.