Modular Integrated Construction Detection Algorithm for Optimized YOLOv8

Abstract Detecting module components at the factory is crucial for safety monitoring, quality control, and productivity enhancement. However, traditional detection methods are not cost-effective and also lack real-time performance. To address these challenges, this study proposes an improved YOLOv8 modular integrated construction detection algorithm, introducing the construction of a small object-YOLO that optimizes the YOLOv8 model by replacing the basic module with a new cross-stage partial network fusion module. In contrast, this module uses deformable convolution Networks v2 to handle geometric variations of objects and focus on relevant image regions. Additionally, the Wise-IoU strategy reduces the competitiveness of high-quality anchor boxes and harmful gradients generated by low-quality examples. The Multi-Head self-attention further improves detection accuracy by capturing the relationship between the image and important objects, making it more suitable for the modular integrated construction dataset. As construction images are often taken from a top or bird's-eye view, small objects can be difficult to detect. Therefore, this algorithm introduces a small object algorithm to enhance the model's ability to detect small objects. Experimental results demonstrate that the improved YOLOv8 model that effectively identifies moving objects. Compared to the original YOLOv8 model, the improved model achieves a 4.4% increase in mAP and a 4.3% increase in F1 score while reducing parameters by 54.05% and GFLOPs by 55.39%. The proposed algorithm will be a reference for automatic detection methods of modular integrated construction at the factory.