LKA-ReID:Vehicle Re-Identification with Large Kernel Attention

TL;DR

This work tackles vehicle re-identification by addressing high inter-class similarity through a novel attention architecture. It introduces Large Kernel Attention (LKA) to capture long-range dependencies efficiently and Hybrid Channel Attention (HCA) to fuse spatial and channel cues, integrated in a four-branch network. On the VeRi-776 dataset, the method achieves strong results with mAP up to 86.65% and Rank-1 up to 98.03%, with ablation analyses showing additive gains from LKA and HCA and parameter efficiency. The approach reduces reliance on extra annotations and demonstrates the practical viability of combining global-local and channel-spatial attention for robust cross-camera vehicle identification, while acknowledging higher computational cost and aiming for lighter designs in future work.

Abstract

With the rapid development of intelligent transportation systems and the popularity of smart city infrastructure, Vehicle Re-ID technology has become an important research field. The vehicle Re-ID task faces an important challenge, which is the high similarity between different vehicles. Existing methods use additional detection or segmentation models to extract differentiated local features. However, these methods either rely on additional annotations or greatly increase the computational cost. Using attention mechanism to capture global and local features is crucial to solve the challenge of high similarity between classes in vehicle Re-ID tasks. In this paper, we propose LKA-ReID with large kernel attention. Specifically, the large kernel attention (LKA) utilizes the advantages of self-attention and also benefits from the advantages of convolution, which can extract the global and local features of the vehicle more comprehensively. We also introduce hybrid channel attention (HCA) combines channel attention with spatial information, so that the model can better focus on channels and feature regions, and ignore background and other disturbing information. Experiments on VeRi-776 dataset demonstrated the effectiveness of LKA-ReID, with mAP reaches 86.65% and Rank-1 reaches 98.03%.

Figures (4)

• Figure 1: The overall architecture of our proposed LKA-ReID.
• Figure 2: Large Kernel Attention.
• Figure 3: (a) Hybrid Channel Attention; (b) Local Average pooling; (c) Global Average pooling.
• Figure 4: Visualization of activation maps on VeRi-776 dataset.