Bangladeshi Native Vehicle Classification Based on Transfer Learning with Deep Convolutional Neural Network.

Area of Science:

• Computer Science
• Artificial Intelligence
• Transportation Engineering

Background:

• Intelligent transportation systems (ITS) require accurate vehicle type classification.
• Deep learning (DL) has shown significant advancements in image classification tasks.

Purpose of the Study:

• To develop and evaluate a DL model for classifying native Bangladeshi vehicle types.
• To improve the accuracy of vehicle recognition in the context of Bangladesh's unique vehicle landscape.

Main Methods:

• A dataset of 10,440 images across 13 Bangladeshi vehicle classes was created.
• Transfer learning using a ResNet-50 architecture with added classification blocks was employed.
• Data augmentation techniques were incorporated to enhance model robustness.

Main Results:

• The proposed ResNet-50 model achieved a classification accuracy of 98.00%.
• The model demonstrated superior performance compared to baseline methods and pre-trained models like AlexNet and VGG-16.
• Evaluation metrics included accuracy, precision, recall, and F1-Score.

Conclusions:

• The developed DL model is highly effective for Bangladeshi vehicle type classification.
• The approach shows promise for enhancing ITS in Bangladesh.
• The method successfully handles variations in vehicle physical properties.