This paper describes the proposed methodology, data used and the results of our participation in the ChallengeTrack 2 (Expr Challenge Track) of the Affective Behavior Analysis in-the-wild (ABAW) Competition 2020. In this competition, we have used a proposed deep convolutional neural network (CNN) model to perform automatic facial expression recognition (AFER) on the given dataset. Our proposed model has achieved an accuracy of 50.77% and an F1 score of 29.16% on the validation set.