MAND: Modality-Aware Novelty Detection for Open-World Egocentric Activity Recognition
Published in ICML 2026 Workshop on Epistemic Intelligence in Machine Learning, 2026
Authors: Hyejeong Im*, Wonseon Lim*, Dae-Won Kim (* Equal contribution)
A modality-aware novelty detection method for open-world egocentric activity recognition. The model leverages multimodal signals to identify previously unseen activities while maintaining strong recognition performance on known classes.
Recommended citation: H. Im*, W. Lim*, D. Kim. (2026). "MAND: Modality-Aware Novelty Detection for Open-World Egocentric Activity Recognition." ICML 2026 Workshop on Epistemic Intelligence in Machine Learning. (* Equal contribution)