Steel defects are a frequent problem in steel companies. Proper quality control can reduce quality problems arising from steel defects. Nowadays, steel defects can detect by automation methods that utilize certain algorithms. Deep learning can help the steel defect detection algorithm become more sophisticated. In this study, we use deep learning CNN with Xception architecture to detect steel defects from images taken from high-frequency and high-resolution cameras. There are two techniques used, and both produce respectively 0.94% and 0.85% accuracy. The Xception architecture used in this case shows optimal and stable performance in the process and its results.