Multi-stage deep learning technique for improving traffic sign recognition

The automatic detection and interpretation of all roadside signs with text information is a very challenging and difficult problem. The main challenge is to distinguish signs with important information which are similar in shapes, sizes, and colors. In this research, we offer a multi-stage deep learning-based approach combined with Optical Character Recognition (OCR) for automatic identification and recognition of text in speed limit traffic signs. The proposed technique consists of novel concept in which deep learning detection and OCR interpretation of attributes is introduced. The technique has been evaluated on real world dataset. A comparison of the data obtained utilizing the proposed approach revealed a significant improvement in speed sign recognition accuracy and misclassifications.