Class Novelty Detection

Class novelty detection aims to identify data points that deviate significantly from a known class of "normal" data, without requiring examples of the novelties during training. Current research focuses on improving the performance and robustness of generative models, such as variational autoencoders (VAEs), and discriminative models, often incorporating techniques like adversarial training and knowledge distillation to enhance anomaly detection capabilities. These advancements are crucial for various applications, including anomaly detection in images and videos, medical diagnosis, and network security, where identifying unseen or unexpected events is critical. The development of efficient and accurate novelty detection methods is driving progress in these fields.