MMF: A Loss Extension for Feature Learning in Open Set Recognition

The objective of open set recognition (OSR) is to classify the known classes as well unknown when collected samples cannot exhaust all classes. This paper proposes a loss extension that emphasizes features with larger and smaller magnitudes find representations can more effectively separate from Our contributions include: First, we introduce an be incorporated into different functions discriminative representations. Second, show proposed significantly improve performances two types on datasets domains. Third, extension, one function outperforms others in training time model accuracy.