Detecting Novelties with Empty Classes
This addresses the challenge of adapting AI systems to evolving environments for applications like autonomous systems or robotics, though it is incremental as it builds on existing anomaly detection methods.
The paper tackles the problem of enabling deep neural networks to detect and learn novel classes in open-world applications without ground truth labels, by using anomaly detection to identify out-of-distribution data and fine-tuning with empty classes and new loss functions, achieving results in image classification and semantic segmentation.
For open world applications, deep neural networks (DNNs) need to be aware of previously unseen data and adaptable to evolving environments. Furthermore, it is desirable to detect and learn novel classes which are not included in the DNNs underlying set of semantic classes in an unsupervised fashion. The method proposed in this article builds upon anomaly detection to retrieve out-of-distribution (OoD) data as candidates for new classes. We thereafter extend the DNN by $k$ empty classes and fine-tune it on the OoD data samples. To this end, we introduce two loss functions, which 1) entice the DNN to assign OoD samples to the empty classes and 2) to minimize the inner-class feature distances between them. Thus, instead of ground truth which contains labels for the different novel classes, the DNN obtains a single OoD label together with a distance matrix, which is computed in advance. We perform several experiments for image classification and semantic segmentation, which demonstrate that a DNN can extend its own semantic space by multiple classes without having access to ground truth.