LatentGNN: Learning Efficient Non-local Relations for Visual Recognition

Image credit: Songyang Zhang

Abstract

Capturing long-range dependencies in feature representations is crucial for many visual recognition tasks. Despite recent successes of deep convolutional networks, it remains challenging to model non-local context relations between visual features. A promising strategy is to model the feature context by a fully-connected graph neural network(GNN), which augments traditional convolutional features with an estimated non-local context representation. However, most GNN-based approaches require computing a dense graph affinity matrix and hence have difficulty in scaling up to tackle complex real-world visual problems. In this work, we propose an efficient and yet flexible non-local relation representation based on a novel class of graph neural networks. Our key idea is to introduce a latent space to reduce the complexity of graph, which allows us to use a low-rank representation for the graph affinity matrix and to achieve a linear complexity in computation. Extensive experimental evaluations on three major visual recognition tasks show that our method outperforms the prior works with a large margin while maintaining a low computation cost.

Publication
In International Conference on Machine Learning, 2019
Songyang Zhang
Songyang Zhang
Shanghai AI Lab

My research interests include few/low-shot learning, graph neural networks and video understanding.

Shipeng Yan
Shipeng Yan
Bytedance

My research interests include few/low-shot learning, incremental learning and representation learning.

Xuming He
Xuming He
Associate Professor

My research interests include few/low-shot learning, graph neural networks and video understanding.

Related