Cognex Deep Learning can quickly identify pores under the same lighting conditions, while other methods are still trying to detect them. Engineers can use a set of "good" and "bad" cylinder images to train the software in a few minutes, using the mask filter program to adjust the region of interest and eliminate the bright circular plane with a hollow center. Technicians can use the Cognex Deep Learning defect detection tool in supervised mode to annotate the pores in the "bad" images and adjust the parameters, including feature size, aspect ratio, aspect ratio and shear modulus, to ensure that the created reference model takes into account all appearance changes. "Good" images that describe normal cylinders can help the software learn what kind of minor casting anomalies and variations are acceptable. Engineers can retrain the system, adjust parameters and add additional images until the model can summarize the normal appearance of the cylinder and identify abnormal conditions. At runtime, the deep learning-based software can detect each image in milliseconds, depicting features with pores as defects and other features as normal.