Human Action Recognition using 3D Convolutional Neural Networks with 3D Motion Cuboids in Surveillance Videos

Abstract In recent days, suspicious action recognition is a significant topic in intelligent video surveillance and computer vision research. Action recognition methodologies are specially needed for surveillance systems which are required to prevent crimes and treacherous actions before occurring. In this paper, we present 3D - Convolutional Neural Networks (3D-CNN) with 3D motion cuboid for action detection and recognizing in videos. The experiments are conducted on benchmark KTH and Weizmann dataset. The proposed method is compared with the existing methods in terms of accuracy. The results show that this approach is outperforms previously published results.