Weighted Local Feature Vehicle Re-identification Network

Abstract

With the rapid development of science and technology, how to accurately identify the same vehicle under different cameras is of great significance to smart city construction. At present, most of the vehicle re-identification methods only use global features, and often neglect the local features that often play an important role in it. To overcome this problem, we propose a multi-scale feature network with an attention module to integrate global and local features. Multi-scale feature fusion to reduce the loss of information caused by network deepening obtained more feature information, and enables the network to learn multi-level feature information. The attention module can make the network pay more attention to the discriminative features of the vehicle, such as windshield stickers and scratches on the vehicle. At the same time, we weighted the local features considerations. Extensive experiments demonstrate the effectiveness of our approach and we achieve state-of-the-art results on two challenging datasets, including VeRi-776 [1-3] and VRIC [4].