<![CDATA[Road damage is a serious issue that can impede traffic and increase the risk of accidents in any area. Fast and accurate detection and classification of road damage are crucial for efficient maintenance and repair. Considering the ease of access, the implementation of this detection can be done using a mobile application. This study aims to compare the performance of two object detection algorithms, YOLOv8 and SSD-MobileNet, in detecting and classifying road damage in mobile application. Evaluation is conducted using accuracy, speed, and memory utilization, and classification of road damage into six categories namely block cracks, alligator cracks, transverse cracks, edge cracks, patches, and potholes using a confusion matrix. The results show that YOLOv8 has an overall accuracy of 86.4%, a speed of 0.5 ms, and consumes 0.41 GB of RAM. SSD-MobileNet shows an overall accuracy of 91.1%, speed 0.7 ms, and consumes 0.14 GB of RAM. The comparison indicates that YOLOv8 excels in detection speed, while SSD-MobileNet is more higher accuracy and efficient in memory. This study is limited to a performance measurement approach for YOLOv8 and SSD-MobileNet algorithms in a mobile-based road defect detection context. Its contribution lies in the trade-off between accuracy, speed, and the memory required to implement the models in limited devices. In future research is recommended to explore model with pruning to reduce memory usage.]]>
Copyrights © 2025
