Deep Learning Based Methods in Image Analytics for Vehicle Detection: A Review
Pramod Vishwakarma, Nitin Jain
Abstract
The literature review reveals a substantial body of work dedicated to the detection, categorization, and tracking of vehicles based on deep learning methods. There has been a lot of recent interest in the research of vehicle detection. Many multinational, collaborative academic-industrial research projects aim to better monitor and control traffic, making it safer and reducing the likelihood of accidents. Among the many sensors used in traffic management, detecting vehicles is essential (ITS). It’s also common knowledge that improving traveller security necessitates a more robust and sophisticated intelligent transportation system. Vehicle detection, tracking, and classification are the three primary operations of a vision-based traffic monitoring and surveillance system. The improved network prioritises the real time detection and also it takes detection accuracy into consideration, doing so through group normalisation to mitigate the effect of batch data size on model accuracy and the method of softening no maximum suppression to cut down on missed detections. The issue of real-time vehicle identification is being studied and researched as part of the bigger topic of intelligent transportation. Some research have offered a concept for the scheme, however there is a major problem: detecting vehicles in real time. In order to build a a scheme for vehicle tracking which is accurate and real time both in accordance with each actual circumstance, the YOLO study suggests a design base employing a deep learning algorithm as the system.