Human Action Recognition and Prediction: A Survey

Derived from rapid advances in computer vision and machine learning, video analysis tasks have been moving inferring the present state to predicting future state. Vision-based action recognition prediction videos are such tasks, where is infer human actions (present state) based upon complete executions, predict (future incomplete executions. These two become particularly prevalent topics recently because of their explosively emerging real-world applications, as visual surveillance, autonomous driving vehicle, entertainment, retrieval, etc. Many attempts devoted last a few decades order build robust effective framework for prediction. In this paper, we survey state-of-the-art techniques Existing models, popular algorithms, technical difficulties, databases, evaluation protocols, promising directions also provided with systematic discussions.