AiATrack: Attention in Attention for Transformer Visual Tracking

TL;DR

AiATrack introduces an attention-in-attention module that improves correlation accuracy in Transformer visual tracking, achieving state-of-the-art results with real-time performance.

Contribution

The paper proposes the AiA module for better correlation in attention mechanisms and a streamlined AiATrack framework utilizing feature reuse and embeddings.

Findings

Achieves state-of-the-art performance on six benchmarks.

Operates in real-time with efficient feature reuse.

Enhances correlation accuracy in attention mechanisms.

Abstract

Transformer trackers have achieved impressive advancements recently, where the attention mechanism plays an important role. However, the independent correlation computation in the attention mechanism could result in noisy and ambiguous attention weights, which inhibits further performance improvement. To address this issue, we propose an attention in attention (AiA) module, which enhances appropriate correlations and suppresses erroneous ones by seeking consensus among all correlation vectors. Our AiA module can be readily applied to both self-attention blocks and cross-attention blocks to facilitate feature aggregation and information propagation for visual tracking. Moreover, we propose a streamlined Transformer tracking framework, dubbed AiATrack, by introducing efficient feature reuse and target-background embeddings to make full use of temporal references. Experiments show that our tracker achieves state-of-the-art performance on six tracking benchmarks while running at a real-time speed.