Demystifying Catastrophic Forgetting in Two-Stage Incremental Object Detector

Qirui Wu, Shizhou Zhang, De Cheng, Yinghui Xing, Di Xu, PENG WANG, Yanning Zhang

TL;DR: We identify the RoI head classifier as the main source of forgetting in two-stage detectors and propose NSGP-RePRE, which uses Regional Prototype Replay and Null Space Gradient Projection to address it.

Abstract: Catastrophic forgetting is a critical chanllenge for incremental object detection (IOD). Most existing methods treat the detector monolithically, relying on instance replay or knowledge distillation without analyzing component-specific forgetting. Through dissection of Faster R-CNN, we reveal a key insight: Catastrophic forgetting is predominantly localized to the RoI Head classifier, while regressors retain robustness across incremental stages. This finding challenges conventional assumptions, motivating us to develop a framework termed NSGP-RePRE. Regional Prototype Replay (RePRE) mitigates classifier forgetting via replay of two types of prototypes: coarse prototypes represent class-wise semantic centers of RoI features, while fine-grained prototypes model intra-class variations. Null Space Gradient Projection (NSGP) is further introduced to eliminate prototype-feature misalignment by updating the feature extractor in directions orthogonal to subspace of old inputs via gradient projection, aligning RePRE with incremental learning dynamics. Our simple yet effective design allows NSGP-RePRE to achieve state-of-the-art performance on the Pascal VOC and MS COCO datasets under various settings. Our work not only advances IOD methodology but also provide pivotal insights for catastrophic forgetting mitigation in IOD. Code will be available soon.

Lay Summary: In this paper, we thoroughly examine Faster R-CNN and find that the main cause of forgetting lies in the classifier within the RoI head. To solve this, we introduce NSGP-RePRE, a method that replays regional prototypes and uses a gradient projection technique to stabilize feature learning. Our approach outperforms existing methods on several benchmarks and offers valuable direction for improving incremental object detection.

Application-Driven Machine Learning: This submission is on Application-Driven Machine Learning.

Link To Code: https://github.com/fanrena/NSGP-RePRE

Primary Area: Applications->Computer Vision

Keywords: Object Detection, Continual Learning

Submission Number: 3148