ROAD CONDITION CLASSIFICATION USING CONVOLUTIONAL NEURAL NETWORKS

Abstract

Autonomous driving is an increasingly important theme nowadays. One of the reasons behind this is the evolution of hardware components in the last years, which made possible both research and implementation of much more complex deep learning techniques. An interesting direction in the vast field of autonomous driving is the discrimination of the condition of the road, with respect to weather. This paper presents a supervised learning based approach to road condition classification. Specifically, we take advantage of the power of Convolutional Neural Networks (CNNs) in the context of image classification. We describe several CNN architectures that use state of the art deep learning techniques and compare their performance. In addition to the simple CNN-based learners, we propose a CNN-based ensemble learner able of a better predictive performance compared to the single models.

Keywords

• Autonomous driving, Road condition classification, Supervised learning, Convolutional Neural Networks, Ensemble learner.