Enhancing the Evaluation Performance of Convolutional Neural Networks-Based Vehicle Classification Systems

The rapid advancement of artificial intelligence, particularly deep learning, has spurred its integration into diverse domains. Among these, the application of deep learning in vehicle classification stands out as a vital focus within intelligent transportation systems, road traffic planning, safety warning, and driverless driving. In this research, a comprehensive approach of CNN’s evolution such as LeNet, AlexNet, and other CNN-based vehicle classification techniques are enhanced for a better performance, while also investigating R-CNN, YOLO, and related methods for vehicle detection. Evaluative datasets for these methods are highlighted. Moreover, the enhancement of the computing power of the Graphics Processing Unit (GPU) and the sharp increase in the amount of data, deep learning based on convolutional neural networks has become a research hotspot. To evaluate the accuracy and efficiency of our vehicle detection methods, RCNN and YOLO series datasets were used to evaluate these methods and an the development of CNN as the main line was introduced and use to analyze the vehicle classification methods based on LeNet, AlexNet and other CNNs. This research is pivotal for vehicle tracking, traffic management, and autonomous driving progress. In this landscape, convolutional neural networks (CNNs), self-encoder neural networks, and deep belief networks take prominence, with CNNs evolving as the most dynamic. Notably, CNN’s development has led to refined vehicle classification through models like LeNet, AlexNet, VGG, and GoogLeNet. The paper concludes by outlining prospective research directions in this domain.