构建两个真实场景的增量检测基准，揭示现有方法在适应与记忆间的严重失衡。
RICO: Two Realistic Benchmarks and an In-Depth Analysis for Incremental Learning in Object Detection
- 提出两个真实数据集构建的增量目标检测基准，模拟真实世界变化
- 小量旧数据重放即超越所有现有方法，但仍不如完整训练
- 发现知识蒸馏弱教师、单一模型泛化不足是核心瓶颈，适合关注实际部署的研究者
增量学习（IL）通过逐批训练新数据实现模型更新，兼顾隐私、效率与可扩展性。但现有评估多依赖合成简化基准，难以反映真实性能。为此，我们提出两个真实增量目标检测基准：域变化型RICO（D-RICO）和类扩展型RICO（EC-RICO），均基于14个多样数据集构建，涵盖真实与合成域、不同天气、时段、相机视角、标注策略等复杂条件。实验表明，所有现有方法在适应新数据与保留旧知识方面均表现不佳；仅用少量旧数据重放即优于全部方法，但依然不及完整训练。我们推测此差距源于知识蒸馏中教师模型能力弱、单模型难以应对多样化任务、以及模型可塑性不足。代码将公开。
原文摘要 · Abstract (English)
Incremental Learning (IL) trains models sequentially on new data without full retraining, offering privacy, efficiency, and scalability. IL must balance adaptability to new data with retention of old knowledge. However, evaluations often rely on synthetic, simplified benchmarks, obscuring real-world IL performance. To address this, we introduce two Realistic Incremental Object Detection Benchmarks (RICO): Domain RICO (D-RICO) features domain shifts with a fixed class set, and Expanding-Classes RICO (EC-RICO) integrates new domains and classes per IL step. Built from 14 diverse datasets covering real and synthetic domains, varying conditions (e.g., weather, time of day), camera sensors, perspectives, and labeling policies, both benchmarks capture challenges absent in existing evaluations. Our experiments show that all IL methods underperform in adaptability and retention, while replaying a small amount of previous data already outperforms all methods. However, individual training on the data remains superior. We heuristically attribute this gap to weak teachers in distillation, single models' inability to manage diverse tasks, and insufficient plasticity. Our code will be made publicly available.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。