Adaptive additive classification-based loss for deep metric learning

Istvan Fehervari, Ives Macedo

Recent works have shown that deep metric learning algorithms can benefit from weak supervision from another input modality. This additional modality can be incorporated directly in…