Res-ViT: a vehicle re-identification method based on residual attention vision transformer

Vehicle re-identification is a crucial area in several domains, including intelligent transportation, smart city, and public security. However, existing methods are either contain interdependent and redundant features, or fail to focus on the global context. We propose a novel method, Res-ViT, which improves the Vision Transformer by leveraging residual network. Specifically, we propose residual attention mechanism and re-ranking to improve the result. We evaluate Res-ViT and the results demonstrate that Res-ViT outperforms in terms of performance, and reliability.