Real-Time Traffic Sign Recognition Based on Efficient CNNs in the Wild

Both unmanned vehicles and driver assistance systems require solving the problem of traffic sign recognition. A lot of work has been done in this area, but no approach has been presented to perform the task with high accuracy and high speed under various conditions until now. In this paper, we have designed and implemented a detector by adopting the framework of faster R-convolutional neural networks (CNN) and the structure of MobileNet. Here, color and shape information have been used to refine the localizations of small traffic signs, which are not easy to regress precisely. Finally, an efficient CNN with asymmetric kernels is used to be the classifier of traffic signs. Both the detector and the classifier have been trained on challenging public benchmarks. The results show that the proposed detector can detect all categories of traffic signs. The detector and the classifier proposed here are proved to be superior to the state-of-the-art method. Our code and results are available online.

I. Introduction

The problem of real-time traffic sign recognition in the wild has to be solved to meet the growing need of automatic driving. The traffic sign recognition usually consists of two steps: traffic sign detection and traffic sign classification. A lot of work has been done to complete the task [1]–[20]. Most of them have trained their models on the German Traffic Sign Detection Benchmark (GTSDB) database [1] (see Figure 1) and the German Traffic Sign Recognition Benchmark (GTSRB) database [2]. Accuracy and speed are surely the two main requirements in practical applications. Although lots of relevant approaches have been presented in previous work, no one can solve the traffic sign recognition problem very well in conditions of different illumination, motion blur, occlusion and so on. So, more effective and more robust approaches need to be developed. Fig. 1.

Examples of traffic scenes in the GTSDB database.