YOLO-AMC: An Improved YOLO Architecture with Attention Mechanisms for Building Crack Detection

2026-06-11 · Source: Computer Vision and Pattern Recognition · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems · Depth: Expert, quick

Summary

YOLO-AMC is an improved YOLO-based architecture designed to enhance automated crack detection in infrastructure inspection and Structural Health Monitoring (SHM). Built upon YOLOv11, this model removes the original C2PSA module and integrates multiple attention mechanisms, including Global Attention Mechanism (GAM), Residual Convolutional Block Attention Module (Res-CBAM), and Shuffle Attention (SA), into its multi-scale feature fusion layers. Experimental results show YOLO-AMC consistently outperforms baseline models YOLOv11n and YOLOv8n. Specifically, GAM achieved the best performance with mAP@0.5 = 0.9917 and mAP@0.5:0.95 = 0.9506, surpassing YOLOv11 (0.9833 / 0.9112) and YOLOv8 (0.9707 / 0.8921). The model maintains a computational complexity of 7.6 GFLOPs, achieving 110.95 FPS on an NVIDIA RTX 4090 and approximately 5 FPS on a Raspberry Pi 5, demonstrating a favorable balance between accuracy and deployment efficiency.

Key takeaway

For Computer Vision Engineers developing automated inspection systems, YOLO-AMC offers a robust solution for challenging crack detection tasks. Your projects requiring high accuracy on low-contrast features, especially for Structural Health Monitoring, can benefit from integrating attention mechanisms like GAM into YOLOv11-based models. This approach provides a strong balance of performance and efficiency, making it suitable for deployment on both high-end GPUs and edge devices like the Raspberry Pi 5.

Key insights

Integrating attention mechanisms into YOLO's feature fusion significantly improves crack detection accuracy and deployment efficiency.

Principles

• Attention mechanisms enhance cross-scale feature integration.
• GAM offers superior performance for low-contrast object detection.
• Edge device deployment requires optimized model architectures.

Method

YOLO-AMC improves YOLOv11 by replacing C2PSA and embedding GAM, Res-CBAM, or SA into Neck's multi-scale feature fusion layers.

In practice

• Apply GAM in YOLO architectures for improved low-contrast object detection.
• Consider YOLO-AMC for real-time structural health monitoring on edge devices.

Topics

• Building Crack Detection
• YOLO Architecture
• Attention Mechanisms
• Structural Health Monitoring
• Edge AI
• Computer Vision

Code references

• CY-Tsai24/YOLO-AMC

Best for: Research Scientist, AI Scientist, Machine Learning Engineer, Computer Vision Engineer

Related on AIssential

• Attention-Guided Autoencoder Fusion for Insulator Defect Detection Using UAV Transmission-Line Imaging — Integrating autoencoders and attention into YOLO enhances UAV-based insulator defect detection, especially for small, rare anomalies.
• Human Activity Recognition Method for Moderate Violence Detection — Skeletal keypoint analysis with Random Forest can detect moderate violence in real-time surveillance.
• YOLOv26: An Object Detector Built for Real-Time Deployment — YOLOv26 prioritizes edge deployment through architectural simplification and hardware-friendly optimizations.
• Small Model, Big Results: How a 5MB On-Device NSFW Detector Outperforms Cloud APIs and Passes a… — Small, on-device object detection models can surpass large cloud APIs in NSFW accuracy and fairness by focusing on anatomical shapes.
• HYolo: An Intelligent IoT-Based Object Detection System Using Hypergraph Learning — HYolo integrates hypergraph learning into YOLO to model complex high-order relationships, improving object detection in IoT.
• Balancing Real and Synthetic Data for CNN-based Masonry Crack Detection — Synthetic data can significantly reduce real data collection needs for CNN-based crack detection while maintaining or improving accuracy.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computer Vision and Pattern Recognition.