Visual Storytelling

TL;DR

This paper introduces the first dataset for sequential vision-to-language tasks, enabling advances in visual storytelling by providing aligned images and narrative data, along with baseline models and evaluation metrics.

Contribution

The paper presents SIND v.1, a novel large-scale dataset for visual storytelling, along with baseline models and an automatic metric for benchmarking progress.

Findings

Established strong baseline models for visual storytelling.

Proposed an automatic metric for evaluating storytelling quality.

Demonstrated potential for AI to understand complex visual narratives.

Abstract

We introduce the first dataset for sequential vision-to-language, and explore how this data may be used for the task of visual storytelling. The first release of this dataset, SIND v.1, includes 81,743 unique photos in 20,211 sequences, aligned to both descriptive (caption) and story language. We establish several strong baselines for the storytelling task, and motivate an automatic metric to benchmark progress. Modelling concrete description as well as figurative and social language, as provided in this dataset and the storytelling task, has the potential to move artificial intelligence from basic understandings of typical visual scenes towards more and more human-like understanding of grounded event structure and subjective expression.