Humans in groups: The importance of contextual information for understanding collective activities
作者：
Highlights：
• People spatial orientation is classified using semisupervised learning.
• A set of descriptors are used to model the context of each individual (people group).
• A complete weighted graph is used to organize descriptors of all members of a group.
• SVMs and a novel kernel between graphs are used to achieve activity classification.
• Results show improvements with respect to previous works.
摘要
Highlights•People spatial orientation is classified using semisupervised learning.•A set of descriptors are used to model the context of each individual (people group).•A complete weighted graph is used to organize descriptors of all members of a group.•SVMs and a novel kernel between graphs are used to achieve activity classification.•Results show improvements with respect to previous works.
论文关键词：Collective activity recognition,People spatial orientation classification,Context-aware people description,Graph kernel,Semi-supervised learning
论文评审过程：Received 1 October 2013, Revised 29 April 2014, Accepted 12 May 2014, Available online 22 May 2014.
论文官网地址：https://doi.org/10.1016/j.patcog.2014.05.008
