Identify Sports Activity from Video Using LSTM

Abstract

Identifying human activity or action is difficult for any autonomous system. But the application of human activity recognition system is limitless. Like, an assistive robot can help people by identifying human activity. But identifying human activity is not easy task so in this work we tried to detect human activity more accurately with combination of CNN and LSTM deep learning algorithms with a large human activity dataset. In this work we first extract feature using CNN algorithm then we will use those features to train a LSTM model for sequence learning.