An Enhanced Lightweight Network for Road Damage Detection Based on Deep Learning

Abstract

Achieving accurate and efficient detection of road damage in complex scenes has always been a challenging task. In this paper, an enhanced lightweight network, E-EfficientDet, is proposed. Firstly, a feature extraction enhancement module (FEEM) is designed to increase the receptive field and improve the feature expression capability of the network, which can extract richer multi-scale feature information. Secondly, to promote the reuse of feature information between different layers in the network and take full advantage of multi-scale context information, four pyramid modules with different structures are designed based on the idea of semi-dense connection, among which the bidirectional feature pyramid network with longitudinal connection (LC-BiFPN) is more suitable for road damage detection. Finally, to meet the road damage detection tasks under different hardware resource constraints, the E-EfficientDet-D0~D2 networks are proposed in this paper based on the compound scaling strategy. Experimental results show that the detection accuracy of E-EfficientDet-D0 improves by 2.41% compared with the original EfficientDet-D0 on the publicly available road damage dataset and outperforms other networks such as YOLOv5s, YOLOv7-tiny, YOLOv4-tiny, Faster R-CNN, and SSD. Meanwhile, the detection speed of EfficientDet-D0 can reach 27.0 FPS, which meets the demand for real-time detection, and the model size is only 32.31 MB, which is suitable for deployment in mobile devices such as unmanned inspection carts, UAVs, and smartphones. In addition, the detection accuracy of E-EfficientDet-D2 can reach 57.51%, which is 4.39% higher than E-EfficientDet-D0, and the model size is 61.78 MB, which is suitable for practical application scenarios that require higher detection accuracy and better hardware performance.

Development of Smart and Lean Pick-and-Place System Using EfficientDet-Lite for Custom Dataset

“…Using Faster-RCNN, Leung [1] developed a vehicle detection approach for insufficient and night-time illumination conditions and improved the mAP values by 0.2; however, it was not meant for a resource-constrained embedded system. Luo [9] achieved an average detection accuracy of 57.51% on EfficientDet-D2 for the road damage detection, which was lower than our 74.1% AP for the Alpha1 dataset. Jain [10] developed "DeepSeaNet" to detect underwater objects with EfficientDet with a high accuracy of 98.63%, but the method was not suitable for lightweight devices due to the complexity of the model.…”

A Review of Deep Learning Advancements in Road Analysis for Autonomous Driving

“…In the same manner of constructing a lightweight model, the authors from [38] introduce E-EfficientDet, a model based on EfficientDet [27]. They integrate a feature extraction enhancement module, utilizing semi-dense connectivity in its feature pyramid, focusing on improving multi-scale feature extraction and information reuse across different network layers.…”

The reading recognition algorithm of pointer meter in unmanned pumping station based on YOLOv5

“…While manual feature extraction may be simple and effective for specific simple tasks, it is not generalisable. In addition, traditional machine learning methods design and optimize feature extraction algorithms and classification algorithms separately, and from a global perspective, the superposition of locally optimal solutions does not necessarily lead to a global optimal solution [6] . With the rapid development of deep learning, deep learning-based target detection algorithms for dial regions offer a new solution to the problem of automatic reading of pointer-type meters.…”