Res3ATN - Deep 3D Residual Attention Network for Hand Gesture Recognition in Videos

Abstract

Hand gesture recognition is a strenuous task to solve in videos. In this\npaper, we use a 3D residual attention network which is trained end to end for\nhand gesture recognition. Based on the stacked multiple attention blocks, we\nbuild a 3D network which generates different features at each attention block.\nOur 3D attention based residual network (Res3ATN) can be built and extended to\nvery deep layers. Using this network, an extensive analysis is performed on\nother 3D networks based on three publicly available datasets. The Res3ATN\nnetwork performance is compared to C3D, ResNet-10, and ResNext-101 networks. We\nalso study and evaluate our baseline network with different number of attention\nblocks. The comparison shows that the 3D residual attention network with 3\nattention blocks is robust in attention learning and is able to classify the\ngestures with better accuracy, thus outperforming existing networks.\n