Cut out and Replay: A Simple yet Versatile Strategy for Multi-Label Online Continual Learning

TL;DR

The paper introduces CUTER, a simple strategy that enhances multi-label online continual learning by identifying label-specific regions for better experience replay, effectively addressing forgetting, missing labels, and class imbalance.

Contribution

It proposes a novel label-specific region identification method that improves continual learning performance and can be integrated with existing approaches.

Findings

Outperforms existing methods on multiple benchmarks

Effectively mitigates catastrophic forgetting and class imbalance

Enhances learning with fine-grained supervision signals

Abstract

Multi-Label Online Continual Learning (MOCL) requires models to learn continuously from endless multi-label data streams, facing complex challenges including persistent catastrophic forgetting, potential missing labels, and uncontrollable imbalanced class distributions. While existing MOCL methods attempt to address these challenges through various techniques, \textit{they all overlook label-specific region identifying and feature learning} - a fundamental solution rooted in multi-label learning but challenging to achieve in the online setting with incremental and partial supervision. To this end, we first leverage the inherent structural information of input data to evaluate and verify the innate localization capability of different pre-trained models. Then, we propose CUTER (CUT-out-and-Experience-Replay), a simple yet versatile strategy that provides fine-grained supervision signals by further identifying, strengthening and cutting out label-specific regions for efficient experience replay. It not only enables models to simultaneously address catastrophic forgetting, missing labels, and class imbalance challenges, but also serves as an orthogonal solution that seamlessly integrates with existing approaches. Extensive experiments on multiple multi-label image benchmarks demonstrate the superiority of our proposed method. The code is available at \href{https://github.com/wxr99/Cut-Replay}{https://github.com/wxr99/Cut-Replay}