VisualBERT: A Simple and Performant Baseline for Vision and Language

TL;DR

VisualBERT introduces a simple Transformer-based framework that effectively models vision-and-language tasks, achieving competitive results while demonstrating the ability to ground language elements to image regions without explicit supervision.

Contribution

The paper presents VisualBERT, a straightforward and flexible model that aligns text and image regions using self-attention, with novel pre-training objectives for vision-and-language tasks.

Findings

Outperforms or rivals state-of-the-art models on multiple tasks

Can ground language to image regions without explicit supervision

Sensitive to syntactic relationships in language understanding

Abstract

We propose VisualBERT, a simple and flexible framework for modeling a broad range of vision-and-language tasks. VisualBERT consists of a stack of Transformer layers that implicitly align elements of an input text and regions in an associated input image with self-attention. We further propose two visually-grounded language model objectives for pre-training VisualBERT on image caption data. Experiments on four vision-and-language tasks including VQA, VCR, NLVR2, and Flickr30K show that VisualBERT outperforms or rivals with state-of-the-art models while being significantly simpler. Further analysis demonstrates that VisualBERT can ground elements of language to image regions without any explicit supervision and is even sensitive to syntactic relationships, tracking, for example, associations between verbs and image regions corresponding to their arguments.