YOLO检测器在持续学习中遗忘旧知识？新方法用自蒸馏解决。
Teach YOLO to Remember: A Self-Distillation Approach for Continual Object Detection
- 用自蒸馏让YOLO在增量学习中保留旧知识，避免遗忘。
- 在VOC和COCO数据集上分别提升mAP 2.1%和2.9%。
- 适合做实时目标检测的持续学习应用，尤其是一阶段无锚框模型。
实时目标检测器如YOLO在大规模数据集上经过多轮训练可达到优异性能。但在真实场景中数据逐批到达时，神经网络会因灾难性遗忘而丢失先前学到的知识。现有研究针对持续学习中的类别增量学习（CIL）提出了多种策略，但多数聚焦于两阶段检测器。研究表明，对一阶段无锚框检测器如YOLO而言，传统的无遗忘学习（LwF）可能无效，因其回归输出噪声大，易传递错误知识。本文提出专为基于YOLO的持续目标检测设计的YOLO LwF方法，采用自蒸馏机制。结合回放记忆后，该方法显著缓解遗忘问题。相比已有方法，在VOC和COCO基准上分别实现+2.1%和+2.9%的mAP提升，达到当前最优水平。
原文摘要 · Abstract (English)
Real-time object detectors like YOLO achieve exceptional performance when trained on large datasets for multiple epochs. However, in real-world scenarios where data arrives incrementally, neural networks suffer from catastrophic forgetting, leading to a loss of previously learned knowledge. To address this, prior research has explored strategies for Class Incremental Learning (CIL) in Continual Learning for Object Detection (CLOD), with most approaches focusing on two-stage object detectors. However, existing work suggests that Learning without Forgetting (LwF) may be ineffective for one-stage anchor-free detectors like YOLO due to noisy regression outputs, which risk transferring corrupted knowledge. In this work, we introduce YOLO LwF, a self-distillation approach tailored for YOLO-based continual object detection. We demonstrate that when coupled with a replay memory, YOLO LwF significantly mitigates forgetting. Compared to previous approaches, it achieves state-of-the-art performance, improving mAP by +2.1% and +2.9% on the VOC and COCO benchmarks, respectively.
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。