Structured Attentions for Visual Question Answering

TL;DR

This paper introduces a novel visual attention model for VQA that encodes cross-region relations using a grid-structured CRF, improving performance on multiple datasets.

Contribution

It models visual attention as a multivariate distribution with CRF and integrates inference algorithms as recurrent neural network layers, enhancing relation encoding.

Findings

Outperforms baseline models on CLEVR by 9.5%

Achieves state-of-the-art on VQA with 1.25% improvement

Demonstrates effective encoding of cross-region relations

Abstract

Visual attention, which assigns weights to image regions according to their relevance to a question, is considered as an indispensable part by most Visual Question Answering models. Although the questions may involve complex relations among multiple regions, few attention models can effectively encode such cross-region relations. In this paper, we demonstrate the importance of encoding such relations by showing the limited effective receptive field of ResNet on two datasets, and propose to model the visual attention as a multivariate distribution over a grid-structured Conditional Random Field on image regions. We demonstrate how to convert the iterative inference algorithms, Mean Field and Loopy Belief Propagation, as recurrent layers of an end-to-end neural network. We empirically evaluated our model on 3 datasets, in which it surpasses the best baseline model of the newly released CLEVR dataset by 9.5%, and the best published model on the VQA dataset by 1.25%. Source code is available at https: //github.com/zhuchen03/vqa-sva.