Exemplar-based Learning for Recognition & Annotation of Human Actions

Abstract

Human action recognition is an active research topic in computer vision. It is a challenging task to model various actions, varying with time resolution, visual appearance and others. For each action category, a large collection of similar actions is learned. This requires training a neural network with a large number of videos. Each action is described as a set of similarities between its instances and candidate exemplars. Then the most discriminative video is chosen. The experiment results on a publicly available dataset known as the KTH dataset. The project is expected to separate the human from the background in the video and identify what action is performed by him/her.