Road damage detection and classification using deep neural networks

• DOI: 10.1007/s42452-024-06129-0
• Publication Date: 2024-08-01T15:03:36Z
• Authors: Yiwen Jiang
• Keywords: Benchmark (surveying); Benchmarking; Road map

Abstract In addressing the challenges of enhancing road damage detection efficiency and accuracy, this paper introduces an optimized YOLOv8 model suitable for embedded systems. The significantly enhances precision, recall, mean Average Precision (mAP), achieving 65.7% mAP on RDD2022 dataset, thereby surpassing models such as Faster R-CNN SSD. This advancement is attributed to integration a Deformable Attention Transformer, GSConv-powered slim-neck module, MPDIoU loss function. These innovations not only contribute model's high performance but also set new benchmark in technology, paving way future enhancements field.