Vehicle Make And Model Classification For Overhead Traffic Analytics

Abstract

Vehicle Make and Model Classification (VMMC) has evolved into a significant subject of study due to its importance in numerous Intelligent Transportation Systems (ITS) and corresponding components such as Automated Vehicular Surveillance (AVS). A highly accurate and real-time VMMC system significantly reduces the overhead cost of resources otherwise required. The VMMC problem is a multiclass classification task with a peculiar set of issues and challenges like multiplicity, inter- and intra-make ambiguity among various vehicle makes and models, which need to be solved in an efficient and reliable manner to achieve a highly robust VMMC system. In this thesis, facing the growing importance of make and model classification of vehicles, we present an image dataset with 94 different classes containing 129,000 images of vehicles in Pakistan to advance the corresponding tasks. Extensive experiments conducted using baseline approaches yield superior results for images that were occluded, under low illumination, partial and overhead camera views, available in our VMMC dataset. The approaches presented herewith provide a robust VMMC system for applications in realistic environments.