Detection of DogRobot Interactions in Video Sequences 1,2
Ali Al-Raziqi, Mahesh Venkata Krishna, Joachim Denzler · 2016
This paper propose a novel framework for unsupervised detection of object interactions in video sequences based on dynamic features. The goal of our system is to process videos in an unsupervised manner using Hierarchical Bayesian Topic Models, specifically the Hierarchical Dirichlet Processes (HDP). We inves� tigate how lowlevel features such as optical flow combined with Hierarchical Dirichlet Process (HDP) can help to recognize meaningful interactions between objects in the scene. For example, in videos of animal interaction recordings, kicking ball, standing, moving around etc. The underlying hypothesis that to validate is that interactions in such scenarios are heavily characterized by their 2D spatiotemporal features. Various experiments have been performed on the challenging JARAIBO dataset and first promising results are reported.