Abstract:The You Only Look Once (YOLO) series algorithms have been widely adopted in concrete crack detection, with attention mechanisms frequently being incorporated to enhance recognition accuracy and efficiency. However, existing research is confronted by two primary challenges: the suboptimal performance of attention mechanism modules and the lack of explanation regarding how these mechanisms influence the model's decision-making process to improve accuracy. To address these issues, a novel Dynamic Efficient Channel Attention (DECA) module is proposed in this study, which is designed to enhance the performance of the YOLOv10 model in concrete crack detection, and the effectiveness of this module is visually demonstrated through the application of interpretable analysis algorithms. In this paper, a concrete dataset with a complex background is used. Experimental results indicate that the DECA module significantly improves the model's accuracy in crack localization and the detection of discontinuous cracks, outperforming the existing Efficient Channel Attention (ECA). When compared to the similarly sized YOLOv10n model, the proposed YOLOv10-DECA model demonstrates improvements of 4.40%, 3.06%, 4.48%, and 5.56% in precision, recall, mAP50, and mAP50-95 metrics, respectively. Moreover, even when compared with the larger YOLOv10s model, these performance indicators are increased by 2.00%, 0.04%, 2.27%, and 1.12%, respectively. In terms of speed evaluation, owing to the lightweight design of the DECA module, the YOLOv10-DECA model achieves an inference speed of 78 frames per second, which is 2.5 times faster than YOLOv10s, thereby fully meeting the requirements for real-time detection. These results demonstrate that an optimized balance between accuracy and speed in concrete crack detection tasks has been achieved by the YOLOv10-DECA model. Consequently, this study provides valuable insights for future research and applications in this field.

A Detection Model for Corner Cracks of Continuous Casting Strand Based on Deep Learning

Deep learning algorithm for real-time automatic crack detection, segmentation, qualification

A Stable Lightweight Model for Metal Crack Detection Based on YOLOv5

Crack Detection of Continuous Casting Slab by Evolutionary Topology Backbone Search

An Improved YOLOX Model for Detecting Strip Surface Defects

YOLO-C: a New Surface Defect Detection Model of Steel Plate Based on YOLO Optimisation.

An efficient steel defect detection model based on multi-scale information extraction

A Deep Learning Framework for Predicting Slab Transverse Crack Using Multivariate LSTM-FCN in Continuous Casting

Real-time detection of surface cracking defects for large-sized stamped parts

SSA-YOLO: An Improved YOLO for Hot-Rolled Strip Steel Surface Defect Detection

Defect Detection for Steel Strip Surface Based on SG-YOLOv5

SteelGuard-yolo: Steel Surface Defect Detection Network Based on Improved YOLOv5s

CRGF-YOLO: An Optimized Multi-Scale Feature Fusion Model Based on YOLOv5 for Detection of Steel Surface Defects

A high-speed YOLO detection model for steel surface defects with the channel residual convolution and fusion-distribution

An Adaptive Shunt Model for Steel Defect Detection Based on YOLOX

DES-YOLO: a novel model for real-time detection of casting surface defects

Deep Learning-Based Understanding of Defects in Continuous Casting Product

Research on steel surface defect classification method based on deep learning

Research on Steel Surface Defect Detection Algorithm Based on Improved YOLOv8n

Longitudinal Crack Detection Approach Based on Principal Component Analysis and Support Vector Machine for Slab Continuous Casting

A Novel YOLOv10-DECA Model for Real-Time Detection of Concrete Cracks