Learning Conditioned Graph Structures for Interpretable Visual Question Answering

TL;DR

This paper introduces a graph-based method for Visual Question Answering that learns question-specific image representations, improving interpretability and achieving promising accuracy on the VQA v2 dataset.

Contribution

It proposes a novel graph learner module that creates question-specific image graphs, enhancing interpretability and capturing semantic and spatial relationships.

Findings

Achieved 66.18% accuracy on VQA v2 dataset.

Demonstrated improved interpretability of visual question answering models.

Showed that question-specific graph representations enhance understanding of image-question interactions.

Abstract

Visual Question answering is a challenging problem requiring a combination of concepts from Computer Vision and Natural Language Processing. Most existing approaches use a two streams strategy, computing image and question features that are consequently merged using a variety of techniques. Nonetheless, very few rely on higher level image representations, which can capture semantic and spatial relationships. In this paper, we propose a novel graph-based approach for Visual Question Answering. Our method combines a graph learner module, which learns a question specific graph representation of the input image, with the recent concept of graph convolutions, aiming to learn image representations that capture question specific interactions. We test our approach on the VQA v2 dataset using a simple baseline architecture enhanced by the proposed graph learner module. We obtain promising results with 66.18% accuracy and demonstrate the interpretability of the proposed method. Code can be found at github.com/aimbrain/vqa-project.