What question did this study set out to answer?

This research aims to evaluate the effectiveness of two YOLO object detection models for robotic copper pipe brazing.

April 11, 2026

Performance Comparison of a Multi-class YOLO Model and Class-wise YOLO Model with Decision-level Output Selection for Welding Position Detection in Copper Pipe Brazing

SYSubin Yang HKHwanhee Kang YKYounggon Kim

Puntos clave

This research aims to evaluate the effectiveness of two YOLO object detection models for robotic copper pipe brazing.
Compared a multi-class YOLO model and a class-wise YOLO model with decision-level selection.
Assessed performance in controlled and uncontrolled industrial environments.
Evaluated models on accuracy, inference speed, and localization precision.
Both models achieved high accuracy in controlled conditions.
Model A delivered faster inference and stable performance in varying environments.
Model B exhibited higher localization precision, making it suitable for detailed detection.

Resumen

This study investigates welding position detection for robotic copper pipe brazing using YOLOv11n-OBB object detection models. Two structures were compared, including a single multi-class model (Model A) and a class-wise trained model with decision-level selection (Model B). Both models were evaluated under identical training conditions in controlled and uncontrolled industrial environments. Results confirmed that while both models achieve high accuracy in controlled settings, Model A provided faster inference and more stable performance under environmental variations, whereas Model B displayed higher localization precision. These findings indicate Model A to be a more practical choice for real-time robotic brazing applications.

Preguntar a la IA

Me gusta

Guardar

Ver artículo completo