Resolving Referring Expressions in Images With Labeled Elements

TL;DR

This paper introduces an end-to-end trainable model that combines element information and image features to improve the resolution of referring expressions in images, demonstrated on COCO-based and webpage datasets.

Contribution

The authors propose a novel architecture that integrates labeled element embeddings with image features for better referring expression resolution.

Findings

Improved accuracy over image-only methods in resolving referring expressions.

Effective incorporation of element bounding box information enhances segmentation.

Validated on COCO and custom webpage datasets.

Abstract

Images may have elements containing text and a bounding box associated with them, for example, text identified via optical character recognition on a computer screen image, or a natural image with labeled objects. We present an end-to-end trainable architecture to incorporate the information from these elements and the image to segment/identify the part of the image a natural language expression is referring to. We calculate an embedding for each element and then project it onto the corresponding location (i.e., the associated bounding box) of the image feature map. We show that this architecture gives an improvement in resolving referring expressions, over only using the image, and other methods that incorporate the element information. We demonstrate experimental results on the referring expression datasets based on COCO, and on a webpage image referring expression dataset that we developed.