The system is also able to automatically group an unsorted set of documents and create classes from these. This mode of operation is called Autoclustering. In clustering mode, the classes are generated automatically and optimized for the image classifier.

The clustering task is optimized to avoid any misclassification. This means that more classes than are needed may be created.

This behavior can be influenced using thresholds and other parameters as described below. In addition, in the products that use the image classifier, you can always join automatically generated classes to one class in a later step to retrain the system.