Vehicle Detection and Classification Using Deep Neural Networks

Urbanized large cities encounter significant challenges of manual traffic control due to overpopulation, traffic congestion and land shortage. These cities have initiated implementing intelligent transportation systems by employing automatic and efficient traffic monitoring and management. Due to bad conditions in traffic, many minor incidents occur on streets, e.g., avoiding traffic rules, over-taking tendency, and high-speed and intoxicated driving, overdrive limit, unsafe lane changes on the road, etc. Sometimes these incidences provoke collisions, accidents, crimes, loss of time and human fatalities. For this reason, automatic vehicle detection and classification are essential to predict traffic congestion levels and lane control and enhance road safety and security. This automated process can solve economic, social, and environmental concerns and impact everyday living issues. In this work, automatic vehicle detection and classification have been introduced using various deep neural network frameworks, i.e., VGG16, VGG19 and YOLOv5. We used transfer learning algorithms built on a conventional neural network for native vehicle identification and classification. This research used a Bangladeshi vehicle images dataset containing 9,058 images. Next, we employed the “ImageDataGenerator” Keras data preprocessing and augmentation framework and various deep learning networks, VGG16, VGG19, and YOLOv5. Finally, the proposed VGG16 and VGG19 models' accuracies for ten classes are 75.15% and 79.57%, respectively. The YOLOv5 architecture attained the best performance with 83.02% accuracy for fifteen vehicle classes.