Machine vision technology with YOLOv5 augmentation technique for metal edge and defect detection
Abstract
Abstract This paper presents a development of YOLOv5-based machine vision technology for potential application in intelligent manufacturing. The YOLOv5 machine vision model was developed to classify and to detect five different metal edge images including burrs, chamfer, dent, hole, and sharp. This study introduces a practical system designed for automated quality control by combining the augmentation technique and YOLOv5 deep learning algorithm. By using smart data augmentation and fine-tuning the model, the system achieved impressive results, including a mean Average Precision (mAP) of 0.95, precision of 0.98, recall of 0.97, and an F1-score of 0.97. These metrics were consistent across various edge types, such as burrs, chamfers, dents, holes, and sharp edges.
Fields of Study & Subjects
Industrial Vision Systems and Defect Detection
Advanced Neural Network Applications
Advanced Data and IoT Technologies
Publication details
- Journal / source
- Journal of Physics Conference Series
- Publisher
- IOP Publishing
- Volume
- 3196
- Issue
- 1
- Pages
- 012101–012101
- Publication date
- 1 March 2026