STMS-YOLOv5: A Lightweight Algorithm for Gear Surface Defect Detection

Abstract

Most deep-learning-based object detection algorithms exhibit low speeds and accuracy in gear surface defect detection due to their high computational costs and complex structures. To solve this problem, a lightweight model for gear surface defect detection, namely STMS-YOLOv5, is proposed in this paper. Firstly, the ShuffleNetv2 module is employed as the backbone to reduce the giga floating-point operations per second and the number of parameters. Secondly, transposed convolution upsampling is used to enhance the learning capability of the network. Thirdly, the max efficient channel attention mechanism is embedded in the neck to compensate for the accuracy loss caused by the lightweight backbone. Finally, the SIOU_Loss is adopted as the bounding box regression loss function in the prediction part to speed up the model convergence. Experiments show that STMS-YOLOv5 achieves frames per second of 130.4 and 133.5 on the gear and NEU-DET steel surface defect datasets, respectively. The number of parameters and GFLOPs are reduced by 44.4% and 50.31%, respectively, while the mAP@0.5 reaches 98.6% and 73.5%, respectively. Extensive ablation and comparative experiments validate the effectiveness and generalization capability of the model in industrial defect detection.

MAA-YOLOv8: enhanced steel surface defect detection through multi-head attention mechanism and lightweight feature fusion

“…Although the model achieved a mean Average Precision (mAP) of 85.9%, the employment of the attention module nevertheless imposed a computational burden on the lightweight adaptation network. Yan et al [18] principally employed ShuffleNetv2 as the backbone network, thereby enhancing the efficacy of feature fusion. They augmented detection accuracy by amalgamating information across multiple scales and resolutions, yet this approach escalated the computational complexity and memory consumption.…”

MFF-YOLO: An Accurate Model for Detecting Tunnel Defects Based on Multi-Scale Feature Fusion

“…Then, fuse both layers to obtain the intermediate layer mapping C. Finally, this layer is fed to D 2 for cross-scale connection and weighting to obtain the D 2 layer. The layer output can be expressed as Equations ( 4) and (5).…”

“…However, conventional methods for finding tunnel defects, such as visual inspection and acoustic inspection, have low detection efficiencies and high result error. Given the quick advancement of deep learning technologies in the detection of targets, such as damage detection and localization of bridge deck pavement [ 2 ], crack detection in concrete bridges [ 3 , 4 ], steel surface flaw detection [ 5 , 6 ], and wheel defect detection [ 7 ] across a variety of disciplines in society, science, and engineering, deep learning-based tunnel flaw detection has recently gained the attention of both domestic and international academics. For example, Sjölander et al [ 8 ] summarized the research on the application of optical detection technology and autonomous evaluation methods based on machine learning technology in tunnel lining inspection, as well as the research on digital cameras, laser scanning, fiber optic sensors, and other methods; they also proposed issues with traditional tunnel inspection methods, such as their low efficiency.…”