Real-time Human Activity Recognition Using ResNet and 3D Convolutional Neural Networks

Abstract

In computer vision-based applications, the recognition of human activity is always a standard problem. Nowadays, activity recognition is more possible and accurate due to good development in artificial neural networks like convolutional neural network CNN. In many recent works, the recognition model architecture use CNN and long short-term memory units (LSTM) - attention models to extract spatial and temporal features from the input video. This particular work is related to real-time human activity recognition by Resnet and 3D CNN without the involvement of the LSTM- attention model. Here the 2D Resnet is modified to 3D CNN to achieve better human activity recognition accuracy. The wide range of data information from the kinetics dataset can avoid overfitting issues during the training period. And the combination of Resnet and 3D CNN can enhance the accuracy of recognition. As a consequence, a method for detecting, monitoring, and recognizing real-time human motion has been developed.