Hierarchical Question-Image Co-Attention for Visual Question Answering

TLDR

Recent VQA work uses attention models that produce spatial maps highlighting image regions relevant to answering questions. This paper argues that modeling question attention—“what words to listen to”—is as important as visual attention. The authors propose a hierarchical co‑attention model that jointly reasons about image and question attention using 1‑D convolutional neural networks. The model raises VQA accuracy from 60.3 % to 60.5 % and COCO‑QA from 61.6 % to 63.3 %, with ResNet further boosting VQA to 62.1 % and COCO‑QA to 65.4 %.

Abstract

A number of recent works have proposed attention models for Visual Question Answering (VQA) that generate spatial maps highlighting image regions relevant to answering the question. In this paper, we argue that in addition to modeling "where to look" or visual attention, it is equally important to model "what words to listen to" or question attention. We present a novel co-attention model for VQA that jointly reasons about image and question attention. In addition, our model reasons about the question (and consequently the image via the co-attention mechanism) in a hierarchical fashion via a novel 1-dimensional convolution neural networks (CNN). Our model improves the state-of-the-art on the VQA dataset from 60.3% to 60.5%, and from 61.6% to 63.3% on the COCO-QA dataset. By using ResNet, the performance is further improved to 62.1% for VQA and 65.4% for COCO-QA.