Vehicle Detection and Classification Using a Deep Learning Approach

Abstract

This research proposes a mechanism to detect and classification if a vehicle will be a car or not car that is the mean of vehicle or non-vehicle. The focus of this research vehicle detection and help for traffic violations. The ability to recognize automobiles in traffic scenes enables the analysis of driver behavior as well as the detection of traffic violations and accidents. Due to the variety of vehicle types and weather and light circumstances, detecting and classifying cars is a difficult operation. Feature extraction methods and Neural Networks are used in a number of solutions. Convolutional neural networks, on the other hand, have been shown to be possibly more effective. We describe a CNN that has been trained to categorize and recognize automobiles from diverse angles in this thesis. In addition, the Keras model is employed for data preprocessing. Our dataset has consist of a total of 3,026 images and txt datasets. For image classification, the Convolutional Neural Network (CNN) is used. On the other hand, to compare with CNN, we have used other algorithms like Inception V3 and AlexNet. I evaluated the Precision, Recall, and F1-score for performance. A vehicle detection system's principal work is to locate one or more automobiles in input photos. CNN beat Inception V3, AlexNet, and had adequate accuracy for vehicle detection and classification tasks, according to the results. This model acquired 97.03% accuracy in classifying vehicles through Convolutional Neural Network (CNN).