With deep-learning-based semantic segmentation, trained defect classes can be localized with pixel accuracy. This allows users to, e.g., solve inspection tasks, which previously could not be realized, or only with significant programming effort.
Semantic segmentation assigns a class to each pixel in the image. There is no distinction between different instances of the same class. Typically, a ‘background‘ class is assigned to all pixels which do not belong to a class of interest. By training the model on a sufficient amount of training data, it eventually learns to predict a class for each pixel in the input image. Furthermore, a confidence score is computed for each pixel in the output.