A hybrid convolutional neural network-gradient boosted classifier for vehicle classification

Abstract

Vehicle tracking and classification are used for intelligent transport system to provide data in terms of traffic management, routing, vehicle volume and others. A new approach will be introduced in this paper, a hybrid classifier that would detect vehicles that would be adaptable to Philippine settings. A combination of convolutional neural network and gradient boosted classifier would boost the classifying accuracy. In the discussion, CNN has outperformed other classifier in terms of accuracy while GBC got the highest AUROC and highest accuracy in terms of classifying. Although CNN and GBC is prone to overfitting, the dataset that will be used contains 1 hour of video.