Resumen de Deep Learning Based Approaches for Vehicle Make and Model Recognition
Shridhar Devamane, Trupthi Rao
Object detection is largely used in the area of computer vision and is critical for variety of applications. During the development of half a century, object detection methods have been continuously developed, and generated numerous approaches which obtained promising achievements. At present, the approach of object detection has been largely evolved into two categories which are traditional machine learning methods utilizing varied computer vision techniques and deep learning methods. In spite of this evolution, accurate implementation of Vehicle Make and Model Recognition (VMMR) is exacting owing to alike (kindred) appearance of different models of vehicles. Therefore this paper presents machine as well as deep learning techniques along with transfer learning models for car detection where the classification is generally at the extent of Make, Model and Year. In this paper, firstly the existing techniques centered on traditional machine learning are introduced and summarized. Then, two main schools of deep learning methods, Convolutional Neural Network (CNN) and ResNeXt50 are selected for analysis. At the end, the methods mentioned are briefly compared and discussed.
