An Efficient and Effective Transformer Decoder-Based Framework for Multi-Task Visual Grounding

TL;DR

This paper introduces EEVG, a Transformer Decoder-based framework for multi-task visual grounding that significantly reduces computational costs, enabling more efficient processing of complex scenes and long language expressions.

Contribution

It proposes a novel Transformer Decoder-based approach that scales linearly with language length and employs a parameter-free method to eliminate background tokens, improving efficiency.

Findings

Achieves state-of-the-art results on visual grounding benchmarks.

Demonstrates reduced computational costs compared to traditional Transformer encoders.

Effective in handling complex scenes with lengthy language expressions.

Abstract

Most advanced visual grounding methods rely on Transformers for visual-linguistic feature fusion. However, these Transformer-based approaches encounter a significant drawback: the computational costs escalate quadratically due to the self-attention mechanism in the Transformer Encoder, particularly when dealing with high-resolution images or long context sentences. This quadratic increase in computational burden restricts the applicability of visual grounding to more intricate scenes, such as conversation-based reasoning segmentation, which involves lengthy language expressions. In this paper, we propose an efficient and effective multi-task visual grounding (EEVG) framework based on Transformer Decoder to address this issue, which reduces the cost in both language and visual aspects. In the language aspect, we employ the Transformer Decoder to fuse visual and linguistic features, where linguistic features are input as memory and visual features as queries. This allows fusion to scale linearly with language expression length. In the visual aspect, we introduce a parameter-free approach to reduce computation by eliminating background visual tokens based on attention scores. We then design a light mask head to directly predict segmentation masks from the remaining sparse feature maps. Extensive results and ablation studies on benchmarks demonstrate the efficiency and effectiveness of our approach. Code is available in https://github.com/chenwei746/EEVG.