Abstract: Real-Time Detection, recognition and audio announcement of traffic sign board is vital and might actually be utilised for driver help to decrease mishaps and ultimately an addon in driverless cars. In this paper Convolution Neural Network (CNN) is utilised to build a framework for Real-Time Recognition of Traffic Signs. The proposed framework works progressively to distinguish and identify Traffic Signs. The commitment of this paper is to an Indian Database of 85 Different Classes of diverse traffic sign boards in the Indian environment. These are gathered from irregular street sides. The pictures were taken from various points and included different boundaries and conditions. A sum of 7210 Images was compiled to form a dataset. CNN engineering was utilised with shifting boundaries to accomplish the best accuracy rates. The test results show that the proposed CNN engineering achieved an accuracy of 99.26% in this manner higher than those achieved in past investigations.

Keywords: Traffic Sign Recognition (TSR), Convolution Neural Network (CNN), Machine Learning, Accuracy and Validation, Open CV

| DOI: 10.17148/IARJSET.2023.10113