Comparative Analysis of Vehicle Make and Model Recognition Using Deep Learning Techniques

Fine-grain vehicle make and model recognition plays a vital role in Intelligent Transport System (ITS). Vehicle Make and Model Recognition has large number of applications in ITS including electronic toll collection, providing visual information for law authority, for various security and authentication purpose. Vision-based fine grained vehicle model recognition is an extremely tough task because of intra-class appearance variation among subordinate-level vehicle classes, with which the fine-grained demonstrate can't be effectively perceived even by a human without domain knowledge. In this way, fine-grained vehicle model recognition needs progressively ground-breaking and discriminative features to classify intra-class objects. This indulges a challenge in this intra-class variety domain, prompting few works of fine-grained vehicle recognition in ITS. Various deep learning strategies can be adequately utilized to defeat this issue of identifying vehicle make and model. Various CNN models with pre-trained weights can be utilized to generate a model which effectively detects if the input image is the front or rear view of the vehicle as well as identify the vehicle make and model utilizing images from both the front as well as rear view.