A Comparative Study of CNN and Transformer Models for Image Recognition in Autonomous Driving

An autonomous driving system requires efficient image recognition to interpret the environment, detect obstacles, and make real-time decisions. This study compares Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs) for image recognition tasks in autonomous driving. CNNs achieve higher accuracy (92.5% vs. 91.2%) and faster inference times (25ms vs. 35ms), making them more suitable for real-time applications. ViTs, however, perform better in challenging conditions, such as low lighting and occlusions, by capturing detailed information across entire images. The findings suggest that both architectures offer distinct advantages and could be combined to enhance the reliability and efficiency of autonomous driving systems.