Wet road detection using CNN with transfer learning

• Main Authors: Mohd Shariff, Khairul Khaizi, MD Ali, Mohd Adli, Enche Ab Rahim, Siti Amlina, Khan Ismail, Zuhani
• Format: Conference or Workshop Item
• Language: English
• Published: Institute of Electrical and Electronics Engineers Inc. 2022
• Subjects: TK7800 Electronics. Computer engineering. Computer hardware. Photoelectronic devices
• Online Access: http://irep.iium.edu.my/98861/7/98861_%20Wet%20road%20detection%20using%20CNN%20with%20transfer%20learning.pdf; http://irep.iium.edu.my/98861/; https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=9794528; https://doi.org/10.1109/ISCAIE54458.2022.9794528
• Institution: Universiti Islam Antarabangsa Malaysia

here is an increasing need to detect wet road surfaces automatically considering many accidents and traffic problems that occur in wet weather. Road condition detection based on acoustic signals has gained more attention in recent years due to its low implementation cost. However, current deep learning methods for wet surface detection rely on supervised audio measurements. Furthermore, they require a large amount of training data. Recent advancements in convolutional neural networks (CNNs) have made it possible for transferring trained CNN from one dataset to another. In this study, we aim to evaluate the capabilities of pre-trained CNN models to detect wet road surfaces. Results show that transfer learning was able to discriminate between dry and wet road surfaces with an accuracy of more than 80%. Additionally, we also provide performance comparisons for the three trained models.