Real-time Road Cracks Detection based on Improved Deep Convolutional Neural Network
Syed Ali Hassan, Seung Heon Han, Soo Young Shin
Abstract
This paper presents road cracks detection implementation to help the road inspectors to easily identify damages on the road. Sometimes due to earthquake cracks appears on the road and identification of damages on the road is required. Identifying road cracks manually during the inspection is a tedious and difficult task. The formation of own data set is a time-consuming and laborious task. Own created data set is labelled with the improvements in the YOLOv3 tiny is finished and compared to better detect the cracks on the road. Both models performance is benchmarked in respect of accuracy and MAP (mean average precision). It is observed in testing phase, the improved tiny version of YOLO performed better in terms of MAP and accuracy. This road cracks detection system can be implemented in UAV or vehicle to detect road cracks using camera vision in Real-Time for the inspection of road.