A Graph-based Approach for Object Detection and Action Recognition in Videos

Abstract

Recently, the field of action and activity recognition in videos has been receiving a lot of attention. However, in order to be able to classify the actions in the scene, it is necessary to first detect the actors or objects involved. In fact, as already pointed out by former studies, there is a direct relation between the quality of the detection and the performance of action recognition methods. In this work we propose a 3-layer framework that incorporates: low-level feature extraction, graph description and action recognition in videos. This framework will be used to perform action recognition and also improve the overall detection of the objects involved.