Generation and Comprehension of Unambiguous Object Descriptions

TL;DR

This paper introduces a deep learning-based method for generating and understanding unambiguous object descriptions in images, outperforming previous approaches and providing a new large-scale dataset for the task.

Contribution

The paper presents a novel deep learning model for unambiguous referring expression generation and comprehension, along with a new large-scale dataset based on MS-COCO.

Findings

Our method outperforms previous approaches in generating unambiguous descriptions.

The dataset enables objective evaluation of referring expression tasks.

The toolbox facilitates visualization and assessment of model performance.

Abstract

We propose a method that can generate an unambiguous description (known as a referring expression) of a specific object or region in an image, and which can also comprehend or interpret such an expression to infer which object is being described. We show that our method outperforms previous methods that generate descriptions of objects without taking into account other potentially ambiguous objects in the scene. Our model is inspired by recent successes of deep learning methods for image captioning, but while image captioning is difficult to evaluate, our task allows for easy objective evaluation. We also present a new large-scale dataset for referring expressions, based on MS-COCO. We have released the dataset and a toolbox for visualization and evaluation, see https://github.com/mjhucla/Google_Refexp_toolbox