Video Dynamics Detection Using Deep Neural Networks

In recent years, deep neural networks (DNNs) have achieved a remarkable progression in solving many complex problems. DNNs are suitable for dealing with the problems related to time series, such as speech recognition and natural language processing. Video dynamics detection, for instance, is time dependent. Apparently, video dynamics detection needs to utilize the present, previous, and next frames of a given video. If a frame change occurs, it triggers whether a video event happens or not. In this paper, video dynamics detection based on deep learning is implemented and our contributions are to effectively improve the accuracy of video dynamics detection. The contributions of this paper are as follows: 1) increasing the accuracy rate to 96% compared to an FSM-based video dynamics detection in real time; 2) By combining convolutional neural network and recurrent neural network (RNN) together, the training time is greatly reduced as we expected.