Harlequin: Color-driven Generation of Synthetic Data for Referring Expression Comprehension

TL;DR

This paper introduces Harlequin, a synthetic data generation framework for Referring Expression Comprehension that creates large-scale, annotated datasets to improve model training without manual labeling.

Contribution

It presents a novel image synthesis pipeline that generates a large artificial dataset for REC, enhancing training and performance of deep learning models.

Findings

Pre-training on Harlequin improves REC model performance.

Harlequin dataset contains over 1 million queries.

Synthetic data reduces reliance on manual annotations.

Abstract

Referring Expression Comprehension (REC) aims to identify a particular object in a scene by a natural language expression, and is an important topic in visual language understanding. State-of-the-art methods for this task are based on deep learning, which generally requires expensive and manually labeled annotations. Some works tackle the problem with limited-supervision learning or relying on Large Vision and Language Models. However, the development of techniques to synthesize labeled data is overlooked. In this paper, we propose a novel framework that generates artificial data for the REC task, taking into account both textual and visual modalities. At first, our pipeline processes existing data to create variations in the annotations. Then, it generates an image using altered annotations as guidance. The result of this pipeline is a new dataset, called Harlequin, made by more than 1M queries. This approach eliminates manual data collection and annotation, enabling scalability and facilitating arbitrary complexity. We pre-train three REC models on Harlequin, then fine-tuned and evaluated on human-annotated datasets. Our experiments show that the pre-training on artificial data is beneficial for performance.