Attributes Guided Feature Learning for Vehicle Re-Identification

Vehicle Re-ID has recently attracted enthusiastic attention due to its potential applications in smart city and urban surveillance. However, it suffers from large intra-class variation caused by view variations illumination changes, inter-class similarity especially for different identities with a similar appearance. To handle these issues, this paper, we propose novel deep network architecture, which guided meaningful attributes including camera views, vehicle types colors Re-ID. In particular, our is end-to-end trained contains three subnetworks of features embedded the corresponding attributes. For training, annotate labels on VeRi-776 dataset. Note that one can directly adopt pre-trained (as well as type color) subnetwork other datasets only ID information, demonstrates generalization model. Extensive experiments benchmark VehicleID suggest proposed approach achieves promising performance yields new state-of-the-art