Comparative Study on Real-Time Vehicle Classification

Vehicle detection and classification play an important role in various fields such as traffic management and traffic surveillance systems. In this paper, we have compared the two most popular machine learning models, SVM, and YOLO for the detection and classification of vehicles on the same dataset which was collected from the GTI vehicle database. SVM(Support Vector Machine) is employed with HoG(Histogram of Gradient) to extract the HoG features and the Sliding window technique to detect the presence of a vehicle within a frame. On the other hand, YOLO, a regression-based algorithm that functions by separating the image into a matrix and looks for predicting classes and bounding boxes rather than interesting regions of the image. After successfully applying both the algorithms we analyzed that YOLO is preferable over SVM for real-time applications as the processing speed of YOLO is 45fps which is much higher than SVM i.e 2fps.