Recently, metal 3D printing technology has developed and has been widely applied in fields such as mechanical parts and construction sites. However, the problem of output defects must be resolved. These defects appear as pores and microcracks in the output, which can be confirmed through microscopic analysis of the output. In addition, if the understanding of pores or cracks is unclear or many images need to be checked in a short time, an error might occur. Therefore, this study aims to develop a precision object detection algorithm using deep learning.
<br>The purpose is to automatically detect defects using deep learning-based You Only Look Once (YOLO). Through comparison using YOLO v3 and v5 algorithms, the accuracy and speed were compared to analyze which YOLO model was efficient in the defect detection process.
