Asphalt Pavement Pothole Detection using Deep learning method based on YOLO Neural Network

• Published in: 2019 International Seminar on Intelligent Technology and Its Applications (ISITIA) pp. 35 - 40
• Main Authors: Ukhwah, Ernin Niswatul, Yuniarno, Eko Mulyanto, Suprapto, Yoyon Kusnendar
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.08.2019
• Subjects: Area measurement; Asphalt; Computer Vision; Distress Detection; Neural networks; Object Detection; Pothole Detection; Roads; Seminars; Surface treatment; YOLO
• DOI: 10.1109/ISITIA.2019.8937176

There is an increasing need for assessment of national road condition. Currently, some automatic devices have been extensively applied to collect up-date data about road condition, such as the use of survey vehicle for collecting data-which make it faster and more accessible, and semi-automatically data processing that is useful for policy decision making. Yet, demand for more detail road data is continuously growing; thus, data improvement needs to perform, upgrading the existing solution. To date, stages on identification and classification of road damages are being conducted semi-manually based on images collected by survey vehicle; it is hindered due to the facts that this method is the cost-consuming process and may result in inconsistency. Therefore, this present work used YOLO with three different architecture configuration, i.e., Yolo v3, Yolo v3 Tiny, and Yolo v3 SPP, enabling us to create a more accurate assessment for detecting potholes on the road surface. The results showed the average mAP values for Yolo v3, Yolo v3 Tiny, and Yolo v3 SPP at 83.43%, 79.33%, and 88.93%. While the area measurement shows the accuracy of 64.45%, 53.26%, and 72.10% respectively. And it needs 0,04 second to detect each image. Conclusively, it shows a satisfying result in pothole detection; thus, this technique has a high opportunity to developed and implemented as a tool for road assessment.