no code implementations • ICCV 2023 • Hyekang Park, Jongyoun Noh, Youngmin Oh, Donghyeon Baek, Bumsub Ham
We present in this paper an in-depth analysis of existing regularization-based methods, providing a better understanding on how they affect to network calibration.
no code implementations • 13 Oct 2022 • Youngmin Oh, Donghyeon Baek, Bumsub Ham
Based on this, we then introduce an adaptive logit regularizer (ALI) that enables our model to better learn new categories, while retaining knowledge for previous ones.
no code implementations • 12 Oct 2022 • Donghyeon Baek, Youngmin Oh, SangHoon Lee, Junghyup Lee, Bumsub Ham
We introduce a CISS framework that alleviates the forgetting problem and facilitates learning novel classes effectively.
Class-Incremental Semantic Segmentation Knowledge Distillation
1 code implementation • 21 Jul 2022 • SangHoon Lee, Youngmin Oh, Donghyeon Baek, Junghyup Lee, Bumsub Ham
To this end, we introduce a novel normalization layer, dubbed ProtoNorm, that calibrates features from pedestrian proposals, while considering a long-tail distribution of person IDs, enabling L2 normalized person representations to be discriminative.
no code implementations • ICCV 2021 • Donghyeon Baek, Youngmin Oh, Bumsub Ham
To this end, we leverage visual and semantic encoders to learn a joint embedding space, where the semantic encoder transforms semantic features to semantic prototypes that act as centers for visual features of corresponding classes.