Open Set

Open-set recognition tackles the challenge of classifying data where unknown classes may be present during testing, unlike traditional closed-set classification. Current research focuses on developing robust models that accurately classify known classes while reliably identifying unknown samples, employing techniques like probabilistic embeddings, adversarial training, and prototype learning within various architectures including vision transformers and convolutional neural networks. This field is crucial for improving the reliability and safety of real-world applications such as autonomous systems, medical diagnosis, and biometric security, where encountering unseen data is inevitable. The development of effective open-set methods is driving advancements in uncertainty estimation and data-centric approaches to improve model generalization and robustness.