Referring Expression Comprehension Using Language Adaptive Inference

TL;DR

This paper introduces LADS, a framework that dynamically extracts language-adaptive subnets from a REC model based on referring expressions, leading to faster and more accurate object localization.

Contribution

The paper proposes a novel dynamic inference framework, LADS, that adapts the REC model to specific expressions, improving efficiency and accuracy over fixed-structure models.

Findings

Achieves faster inference speed

Attains higher accuracy on benchmark datasets

Demonstrates effectiveness of expression-specific subnets

Abstract

Different from universal object detection, referring expression comprehension (REC) aims to locate specific objects referred to by natural language expressions. The expression provides high-level concepts of relevant visual and contextual patterns, which vary significantly with different expressions and account for only a few of those encoded in the REC model. This leads us to a question: do we really need the entire network with a fixed structure for various referring expressions? Ideally, given an expression, only expression-relevant components of the REC model are required. These components should be small in number as each expression only contains very few visual and contextual clues. This paper explores the adaptation between expressions and REC models for dynamic inference. Concretely, we propose a neat yet efficient framework named Language Adaptive Dynamic Subnets (LADS), which can extract language-adaptive subnets from the REC model conditioned on the referring expressions. By using the compact subnet, the inference can be more economical and efficient. Extensive experiments on RefCOCO, RefCOCO+, RefCOCOg, and Referit show that the proposed method achieves faster inference speed and higher accuracy against state-of-the-art approaches.