Automated labeling method for direction detection of objects inside X-ray image
Jungi Lee, Hyun-Yong Lee, Nac‐Woo Kim, Yu-Min Hwang, Seok‐Kap Ko · 2023
To utilize image datasets for deep learning models, high-quality labeling data is required for good performance. In the case of acquiring a dataset, there are difficulties such as making an annotation by manpower. Many studies or software are provided to increase the convenience of such labeling. In this paper, YOLOX-based object detection, Canny edge detection, and Hough transformation-based linear component extraction techniques are applied to automatically label the position of a linear structure inside an X-ray image. To verify this method, actually acquired X-ray images of wind turbine blades were used in this experiment.