Flickr30k Entities: Collecting Region-to-Phrase Correspondences for Richer Image-to-Sentence Models

TLDR

The Flickr30k dataset is a standard benchmark for sentence‑based image description, and its annotation is essential for advancing automatic image description and grounded language understanding. The paper introduces Flickr30k Entities, augmenting Flickr30k with coreference chains and bounding boxes, and demonstrates its usefulness for text‑to‑image reference resolution and bidirectional image‑sentence retrieval. The authors augment Flickr30k with 244 k coreference chains and 276 k bounding boxes, then conduct experiments on text‑to‑image reference resolution and image‑sentence retrieval. The experiments confirm that training with explicit region‑to‑phrase correspondence improves state‑of‑the‑art retrieval accuracy, yet accurately inferring this correspondence from an image and caption remains very challenging.

Abstract

The Flickr30k dataset has become a standard benchmark for sentence-based image description. This paper presents Flickr30k Entities, which augments the 158k captions from Flickr30k with 244k coreference chains linking mentions of the same entities in images, as well as 276k manually annotated bounding boxes corresponding to each entity. Such annotation is essential for continued progress in automatic image description and grounded language understanding. We present experiments demonstrating the usefulness of our annotations for text-to-image reference resolution, or the task of localizing textual entity mentions in an image, and for bidirectional image-sentence retrieval. These experiments confirm that we can further improve the accuracy of state-of-the-art retrieval methods by training with explicit region-to-phrase correspondence, but at the same time, they show that accurately inferring this correspondence given an image and a caption remains really challenging.