Make Graph-based Referring Expression Comprehension Great Again through Expression-guided Dynamic Gating and Regression

TL;DR

This paper improves graph-based referring expression comprehension by introducing a dynamic gating module and a regression strategy, significantly enhancing performance without pretraining and surpassing transformer-based methods.

Contribution

It proposes a novel expression-guided dynamic gating and regression approach that addresses noise and localization issues in graph-based REC methods, achieving state-of-the-art results.

Findings

Outperforms transformer-based methods without pretraining

Effectively reduces irrelevant proposals during reasoning

Consistently boosts performance across multiple datasets

Abstract

One common belief is that with complex models and pre-training on large-scale datasets, transformer-based methods for referring expression comprehension (REC) perform much better than existing graph-based methods. We observe that since most graph-based methods adopt an off-the-shelf detector to locate candidate objects (i.e., regions detected by the object detector), they face two challenges that result in subpar performance: (1) the presence of significant noise caused by numerous irrelevant objects during reasoning, and (2) inaccurate localization outcomes attributed to the provided detector. To address these issues, we introduce a plug-and-adapt module guided by sub-expressions, called dynamic gate constraint (DGC), which can adaptively disable irrelevant proposals and their connections in graphs during reasoning. We further introduce an expression-guided regression strategy (EGR) to refine location prediction. Extensive experimental results on the RefCOCO, RefCOCO+, RefCOCOg, Flickr30K, RefClef, and Ref-reasoning datasets demonstrate the effectiveness of the DGC module and the EGR strategy in consistently boosting the performances of various graph-based REC methods. Without any pretaining, the proposed graph-based method achieves better performance than the state-of-the-art (SOTA) transformer-based methods.