Learning Connectivity with Graph Convolutional Networks for Skeleton-based Action Recognition

TL;DR

This paper introduces a novel GCN framework that learns the graph topology during training, improving skeleton-based action recognition by optimizing both convolutional parameters and the graph structure.

Contribution

The proposed method automatically learns graph topologies within GCNs, enhancing performance over handcrafted graphs in action recognition tasks.

Findings

Outperforms existing methods on skeleton-based action recognition

Automatically learns relevant graph topologies during training

Demonstrates superior results compared to handcrafted graph designs

Abstract

Learning graph convolutional networks (GCNs) is an emerging field which aims at generalizing convolutional operations to arbitrary non-regular domains. In particular, GCNs operating on spatial domains show superior performances compared to spectral ones, however their success is highly dependent on how the topology of input graphs is defined. In this paper, we introduce a novel framework for graph convolutional networks that learns the topological properties of graphs. The design principle of our method is based on the optimization of a constrained objective function which learns not only the usual convolutional parameters in GCNs but also a transformation basis that conveys the most relevant topological relationships in these graphs. Experiments conducted on the challenging task of skeleton-based action recognition shows the superiority of the proposed method compared to handcrafted graph design as well as the related work.