Road damage detection algorithm for improved YOLOv5

Abstract

Road damage detection is an important task to ensure road safety and realize the timely repair of road damage. The previous manual detection methods are low in efficiency and high in cost. To solve this problem, an improved YOLOv5 road damage detection algorithm, MN-YOLOv5, was proposed. We optimized the YOLOv5s model and chose a new backbone feature extraction network MobileNetV3 to replace the basic network of YOLOv5, which greatly reduced the number of parameters and GFLOPs of the model, and reduced the size of the model. At the same time, the coordinate attention lightweight attention module is introduced to help the network locate the target more accurately and improve the target detection accuracy. The KMeans clustering algorithm is used to filter the prior frame to make it more suitable for the dataset and to improve the detection accuracy. To improve the generalization ability of the model, a label smoothing algorithm is introduced. In addition, the structure reparameterization method is used to accelerate model reasoning. The experimental results show that the improved YOLOv5 model proposed in this paper can effectively identify pavement cracks. Compared with the original model, the mAP increased by 2.5%, the F1 score increased by 2.6%, and the model volume was smaller than that of YOLOv5. 1.62 times, the parameter was reduced by 1.66 times, and the GFLOPs were reduced by 1.69 times. This method can provide a reference for the automatic detection method of pavement cracks.

MR-YOLO: An Improved YOLOv5 Network for Detecting Magnetic Ring Surface Defects

“…Xie et al [41] proposed a lightweight metal surface defect detection algorithm EMV2YOLOv4 based on YOLOv4, which could meet the requirements of lightweight deployment and accuracy requirements of metal surface defect detection. Similarly, Guo et al [42] used MobileNetV3 to replace the basic network of YOLOv5 for road damage detection, which greatly reduced the number of parameters and GFLOPs of the model and reduced the size of the model. Combined with the K-Means clustering algorithm, the accuracy and model size are guaranteed.…”

High-resolution processing and sigmoid fusion modules for efficient detection of small objects in an embedded system

“…Various convolutional neural network (CNN)-based object detection methods have recently been proposed in the literature. The well-known networks include a two-stage detector using region proposal called regions with convolutional neural networks (RCNN) 2 and one-stage anchor-based detectors called YOLO series [3][4][5] . However, these methods are not suitable for small object detection since their accuracy is guaranteed only when an object is sufficiently large.…”

A Computer Vision Enabled damage detection model with improved YOLOv5 based on Transformer Prediction Head

“…More recently, a semi-supervised DL-based pixel-level segmentation model (Karaaslan et al, 2021) has been proposed utilizing attention guidance for cracks and spalls localization that reduces computational cost significantly. In separate work, an improved YOLOv5 road damage detection algorithm (Guo and Zhang, 2022) has been proposed leveraging MobileNetV3 as a backbone feature extractor. In Hacıefendioglu and Başaga (2022), Faster R-CNN has been employed for concrete pavement crack detection under various illumination and weather conditions.…”

“…In recent years, various attempts have been made on DL-based computer vision models for damage detection such as Faster R-CNN (Kluger et al, 2018;Wang et al, 2018a), SSD (Maeda et al, 2018;Wang et al, 2018b), RetinaNet (Angulo et al, 2019), YOLO (Alfarrarjeh et al, 2018;Mandal et al, 2020), YOLOv2 (Majidifard et al, 2020), YOLOv5 (Guo and Zhang, 2022) etc. Although the aforementioned techniques have demonstrated outstanding performance, however, the damage detection task faces several challenges, in particular, due to the presence of complex and noisy backgrounds, significant variability of lightening conditions, low visibility, densely packed classes, and overlap, the coexistence of multi-object classes with various aspect ratios, and other morphological characteristics (Azimi et al, 2020;Naddaf-Sh et al, 2020).…”