Connecting Vision and Language with Localized Narratives

TL;DR

This paper introduces Localized Narratives, a novel multimodal annotation method that links detailed language descriptions to specific image regions through synchronized voice and mouse movements, enabling precise visual grounding.

Contribution

It presents a new annotation technique and a large dataset of 849k images with localized descriptions, enhancing multimodal vision-language research.

Findings

Annotations are diverse and accurate.

The method is efficient to produce.

Useful for controlled image captioning.

Abstract

We propose Localized Narratives, a new form of multimodal image annotations connecting vision and language. We ask annotators to describe an image with their voice while simultaneously hovering their mouse over the region they are describing. Since the voice and the mouse pointer are synchronized, we can localize every single word in the description. This dense visual grounding takes the form of a mouse trace segment per word and is unique to our data. We annotated 849k images with Localized Narratives: the whole COCO, Flickr30k, and ADE20K datasets, and 671k images of Open Images, all of which we make publicly available. We provide an extensive analysis of these annotations showing they are diverse, accurate, and efficient to produce. We also demonstrate their utility on the application of controlled image captioning.