Multi-scale multi-stream deep network for car logo recognition

Vehicle logo retrieval or recognition is being a popular study of field in intelligent/smart traffic system. Therefore this paper proposes a multi-scale multi-stream deep network for vehicle logo recognition. The proposed network processes the input vehicle logo image through each of the multi-scale stream to extract the robust features followed by the logo recognition module. The proposed network follows the knowledge sharing strategy by sharing learned features at each stream across the network. We have utilized the existing large-scale benchmark vehicle logo image dataset to validate the proposed network for vehicle logo recognition task. The performance of the proposed network is compared with the existing state-of-the-art deep network which exhibits that the proposed network is superior for the vehicle recognition task.