Deep Learning Based Steel Pipe Weld Defect Detection

Steel pipes are widely used in high-risk and high-pressure scenarios such as oil, chemical, natural gas, shale etc. If there is some defect steel pipes, it will lead to serious adverse consequences. Applying object detection the field of deep learning pipe weld identification can effectively improve inspection efficiency promote development industrial automation. Most predecessors traditional computer vision methods applied detect defects seams. However, rely on prior knowledge only with a single feature, so difficult complete task multi-defect classification, while end-to-end. In this paper, state-of-the-art single-stage algorithm YOLOv5 proposed be compared two-stage representative Faster R-CNN. The experimental results show that applying greatly accuracy, multi-classification task, meet criteria real-time detection.