Image Classification of Heavy Load Vehicles using CNN (Convolutional Neural Network)
Abstract – The most basic desires of road users are for less pollution, less noise, and fewer accidents. Pollution, such as noise and air, is a major source of concern, but vehicle accidents also play a significant impact. Road traffic accidents are the leading cause of death among children and individuals aged 8 to 30. Traffic accidents are one of the most worrisome topics. One of the biggest causes of death and injury is car accidents. Approximately 1.29 million people are killed in automobile accidents each year. Road accidents cost most countries 3% of their GDP. According to the World Health Organization (WHO), speeding, human nuisances (drink and drive), bad road infrastructure, violating traffic restrictions, and incorrectly loaded cars are all factors in accidents. In which improperly loaded cars, either directly or indirectly, cause accidents. These can be reduced if we divide the vehicles into two categories: good load vehicles and bad load cars. As a result, we employed a CNN network with multiple layers, including different type layers, ReLU, pooling layers, dense layers, and so on. We also use batch normalization and dropout layers to prevent the model from becoming overfit. To improve the accuracy of the outcome, we applied augmentation techniques. The effect of employing Max polling in CNN for feature mapping and reducing over fitting is shown below. With a 5 CNN hidden layer model, we achieved 99 percent training accuracy and 91 percent testing accuracy, and with average polling in CNN, we achieved 98 percent training accuracy and 87 percent testing accuracy in both cases. The model's output will aid in predicting the difference between loaded and unloaded cars.
Keywords: Deep Learning, Max Pooling, Artificial Neural Networks, Machine Learning, Support Vector Method, Convolutional Neural Networks, Image Recognition, Image Classification, Augmentation Method