Group Activity Recognition
11 papers with code • 2 benchmarks • 2 datasets
Group Activity Recognition is a subset of human activity recognition problem which focuses on the collective behavior of a group of people, resulted from the individual actions of the persons and their interactions. Collective activity recognition is a basic task for automatic human behavior analysis in many areas like surveillance or sports videos.
To this end, we propose to build a flexible and efficient Actor Relation Graph (ARG) to simultaneously capture the appearance and position relation between actors.
Within each interaction field, we apply DR to predict the relation matrix and DW to predict the dynamic walk offsets in a joint-processing manner, thus forming a person-specific interaction graph.
In group activity recognition, the temporal dynamics of the whole activity can be inferred based on the dynamics of the individual people representing the activity.
In order to model both person-level and group-level dynamics, we present a 2-stage deep temporal model for the group activity recognition problem.
Second, we propose a Relational Autoencoder model for unsupervised learning of features for action and scene retrieval.
We propose to use Normalized cross-correlation (NCC) and the sum of absolute differences (SAD) to calculate the pair-wise appearance similarity and build the actor relationship graph to allow the graph convolution network to learn how to classify group activities.
To understand human behavior we must not just recognize individual actions but model possibly complex group activity and interactions.
Group Activity Recognition Using Joint Learning of Individual Action Recognition and People Grouping
This paper proposes joint learning of individual action recognition and people grouping for improving group activity recognition.