Traffic signs are integral component of road transport infrastructure. They deliver essential driving information to road users, which in turn require them to adapt their driving behavior to road regulations. In this study, the deep learning model is implemented with You Only Look Once (YOLO) based on active learning approach in order to minimize the size of the labeled dataset and provide higher accuracy. This is an efficient approach using a single monocular camera to perform real-time traffic signs recognition and distance estimation between traffic sign and camera pose. YOLO is one of the faster object detection algorithms, and it is a very good choice when real-time detection is needed, without loss of too much accuracy. The active learning algorithm will bridge the gap between labeled and unlabeled data, thus, only queries the samples that would lead to increase the accuracy. The aim of this work is to alarm and notify the drivers without having them to switch their focus. The results of the performed experiments show that about 97% recognition accuracies could be achieved with realtime capability in different real-world scenarios.
|Title of host publication||The 6th International Conference Actual Problems of System and Software Engineering|
|Subtitle of host publication||IEEE|
|Publisher||CEUR Workshop Proceedings (CEUR-WS.org)|
|Publication status||Published - 9 Dec 2019|