期刊
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
卷 44, 期 12, 页码 9209-9216出版社
IEEE COMPUTER SOC
DOI: 10.1109/TPAMI.2021.3124133
关键词
Task analysis; Detectors; Object detection; Training; Proposals; Standards; Feature extraction; Object detection; incremental learning; deep neural networks; meta-learning; gradient preconditioning
资金
- DST through the IMPRINT program [IMP/2019/000250]
- TCS PhD Fellowship
- VR [2016-05543]
- Swedish Research Council [2016-05543] Funding Source: Swedish Research Council
In this study, a meta-learning approach is proposed to achieve incremental learning by learning how to reshape model gradients. Compared to existing methods, this approach can adapt to new tasks and performs well in various incremental learning scenarios.
In a real-world setting, object instances from new classes can be continuously encountered by object detectors. When existing object detectors are applied to such scenarios, their performance on old classes deteriorates significantly. A few efforts have been reported to address this limitation, all of which apply variants of knowledge distillation to avoid catastrophic forgetting. We note that although distillation helps to retain previous learning, it obstructs fast adaptability to new tasks, which is a critical requirement for incremental learning. In this pursuit, we propose a meta-learning approach that learns to reshape model gradients, such that information across incremental tasks is optimally shared. This ensures a seamless information transfer via a meta-learned gradient preconditioning that minimizes forgetting and maximizes knowledge transfer. In comparison to existing meta-learning methods, our approach is task-agnostic, allows incremental addition of new-classes and scales to high-capacity models for object detection. We evaluate our approach on a variety of incremental learning settings defined on PASCAL-VOC and MS COCO datasets, where our approach performs favourably well against state-of-the-art methods. Code and trained models: https://github.com/JosephKJ/iOD.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据