Structure Inference Machines: Recurrent Neural Networks for Analyzing Relations in Group Activity Recognition
Zhiwei Deng, Arash Vahdat, Hexiang Hu, Greg Mori; Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016, pp. 4772-4781
Abstract
Rich semantic relations are important in a variety of visual recognition problems. As a concrete example, group activity recognition involves the interactions and relative spatial relations of a set of people in a scene. State of the art recognition methods center on deep learning approaches for training highly effective, complex classifiers for interpreting images. However, bridging the relatively low-level concepts output by these methods to interpret higher-level compositional scenes remains a challenge. Graphical models are a standard tool for this task. In this paper, we propose a method to integrate graphical models and deep neural networks into a joint framework. Instead of using a traditional inference method, we use a sequential inference modeled by a recurrent neural network. Beyond this, the appropriate structure for inference can be learned by imposing gates on edges between nodes. Empirical results on group activity recognition demonstrate the potential of this model to handle highly structured learning tasks.
Related Material
[pdf]
[
bibtex]
@InProceedings{Deng_2016_CVPR,
author = {Deng, Zhiwei and Vahdat, Arash and Hu, Hexiang and Mori, Greg},
title = {Structure Inference Machines: Recurrent Neural Networks for Analyzing Relations in Group Activity Recognition},
booktitle = {Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2016}
}