Driver Assistant System using YOLO V3 and VGGNET

• Main Authors: Rani, S. Jansi, Eswaran, S. Uthirapathi, Mukund, A.V. Vedha, Vidul, M.
• Format: Conference Proceeding
• Language: English
• Subjects: automated driving assistance; Computational modeling; Computer architecture; Convolutional neural networks; Graphics processing units; Hardware; machine learning; neural networks; Real-time systems; Software; traffic sign recognition; VGGNet; YOLO
• ISSN: 2767-7788
• DOI: 10.1109/ICICT54344.2022.9850535

The ability to observe and comprehend both things that remain static and being non static around the vehicle in a variety of driving and climatic circumstance is the fundamental criteria for autonomous vehicles and advanced systems that assist in driving. Convolutional neural networks (CNN) have the potential to fulfil the current promise of delivering safe driving assistance in modern automobiles. In this study, the performance of YOLO-based software and VGGNet-based architecture for traffic sign detection that is augmented with a CNN is compared, because the detection should be real-time and quick for the driving to be safe, the networks used in this article have only been trained to recognise and classify items such as traffic signs and lights. Detected traffic signs are then forwarded to CNN, where they are classified into one of 43 categories. We exhibit a great degree of classification confidence by correctly recognizing more than 98.9 percent (YOLO) and 96 percent (VGGNet) of evaluated signs in a multitude of environments.