Nonparametric discovery of activity patterns from video collections

Michael C. Hughes, Erik B. Sudderth
2012 2012 IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops  
We propose a nonparametric framework based on the beta process for discovering temporal patterns within a heterogenous video collection. Starting from quantized local motion descriptors, we describe the long-range temporal dynamics of each video via transitions between a set of dynamical behaviors. Bayesian nonparametric statistical methods allow the number of such behaviors and the subset exhibited by each video to be learned without supervision. We extend the earlier beta process HMM in two
more » ... ys: adding data-driven MCMC moves to improve inference on realistic datasets, and using a hierarchical beta process HMM (HBP-HMM) to improve behavior sharing among videos with the same category label. We illustrate discovery of intuitive and useful dynamical structure, at various temporal scales, from videos of simple exercises, Olympic sporting events, and recipe preparation. Video retrieval experiments show that our approach leads to quantitative improvements over conventional bag-of-feature representations.
doi:10.1109/cvprw.2012.6239170 dblp:conf/cvpr/HughesS12 fatcat:xg5bvpi3krelnbx7cgjztautv4