Facial expression recognition in the wild, by fusion of deep learnt and hand-crafted features

Gajjala Viswanatha Reddy, C. V. R. Dharma Savarni, Snehasis Mukherjee

Abstract: Highlights•Two CNNs are trained with different sampling techniques.•Pre-training and fine tuning are used for training the deep networks.•Combined hand crafted features with CNN features to apply in the wild scenario.•Facial landmark points are used to extract the handcrafted features.•Rigorous experiments on Affectnet dataset show superiority of the proposed method.