Classification

Classification is the assignment of an object to one of several categories based on selected features. When working with images, the objects being classified are typically pixels or regions. To assign an object to a specific class, the boundaries between feature spaces separating each class must first be defined via a training procedure. Then, when classifying an unknown object, the class with the largest correspondence between the feature values used for training and the feature values of the unknown object is returned. Often, selecting the features used to create an accurate classifier can be time consuming. To simplify this process, MVTec developed the Automatic Feature Selection tool to automatically choose the best feature set used when building a classifier.

MVTec’s software products support several different classification techniques including:

• Support Vector Machines (SVM)
• k-Nearest Neighbors (kNN)
• Multi-Layer Perceptrons (MLP)
• Gaussian Mixture Models (GMM)
• Convolutional Neural Networks (CNNs)

Classification can be used for:

• Image Segmentation
  to segment images into regions of similar color or texture
• Object Recognition
  to find objects of a specific type within a set of different object types
• Quality Control
  to decide if objects are good or bad
• Novelty Detection
  to detect changes or defects of objects
• Optical Character Recognition (OCR)