Shot Open Set Recognition

Few-shot open-set recognition (FSOR) tackles the challenge of classifying images into known categories and rejecting unknown ones, using only a limited number of labeled examples per known class. Current research focuses on developing robust methods for generating representative negative prototypes of unknown classes, often employing energy-based models, hypernetworks, or adapting existing classification models through techniques like transductive learning. These advancements are crucial for improving the accuracy and reliability of image retrieval systems, anomaly detection, and other applications where dealing with unseen data is essential.