Title Performance Analysis of Temporary Structure Recognition Using YOLOv8 and YOLOv11 Segmentation Models
Authors 우윤희(Woo, Yun-Hee) ; 유무영(Yoo, Moo-Young)
DOI https://doi.org/10.5659/JAIK.2025.41.10.339
Page pp.339-348
ISSN 2733-6247
Keywords Temporary Structures; Object Detection; Segmentation Performance; Smart Annotation
Abstract This study evaluates the detection performance of YOLOv8 and YOLOv11 segmentation models for temporary structures in construction images. The dataset consists of labeled images featuring scaffolding, formwork, and other temporary components commonly found on construction sites. Both models were trained and tested under the same conditions, and their performance was compared using mean Average Precision (mAP) and Frames Per Second (FPS) as evaluation metrics. To analyze the impact of labeling quality on detection results, two annotation methods?manual polygon and smart polygon?were applied. The experimental results showed that the YOLOv8s model achieved the most balanced performance between accuracy and speed, making it suitable for real-time applications. In addition, smart polygon annotations enhanced boundary precision and improved detection accuracy, particularly for complex-shaped structures. This study demonstrates the practical applicability of segmentation-based object detection models in construction environments and provides insight into the development of intelligent visual monitoring systems for automation and quality control in the field.