Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Better Matching, Less Forgetting: A Quality-Guided Matcher for Transformer-based Incremental Object Detection

Qirui Wu, Shizhou Zhang, De Cheng, Yinghui Xing, Lingyan Ran, Dahu Shi, Peng Wang

TL;DR

A Quality-guided Min-Cost Max-Flow (Q-MCMF) matcher, which eliminates harmful supervision from background foregrounding while maximizing foreground learning signals and consistently outperforms existing state-of-the-art approaches.

Abstract

Incremental Object Detection (IOD) aims to continuously learn new object classes without forgetting previously learned ones. A persistent challenge is catastrophic forgetting, primarily attributed to background shift in conventional detectors. While pseudo-labeling mitigates this in dense detectors, we identify a novel, distinct source of forgetting specific to DETR-like architectures: background foregrounding. This arises from the exhaustiveness constraint of the Hungarian matcher, which forcibly assigns every ground truth target to one prediction, even when predictions primarily cover background regions (i.e., low IoU). This erroneous supervision compels the model to misclassify background features as specific foreground classes, disrupting learned representations and accelerating forgetting. To address this, we propose a Quality-guided Min-Cost Max-Flow (Q-MCMF) matcher. To avoid forced assignments, Q-MCMF builds a flow graph and prunes implausible matches based on geometric quality. It then optimizes for the final matching that minimizes cost and maximizes valid assignments. This strategy eliminates harmful supervision from background foregrounding while maximizing foreground learning signals. Extensive experiments on the COCO dataset under various incremental settings demonstrate that our method consistently outperforms existing state-of-the-art approaches.

Better Matching, Less Forgetting: A Quality-Guided Matcher for Transformer-based Incremental Object Detection

TL;DR

A Quality-guided Min-Cost Max-Flow (Q-MCMF) matcher, which eliminates harmful supervision from background foregrounding while maximizing foreground learning signals and consistently outperforms existing state-of-the-art approaches.

Abstract

Incremental Object Detection (IOD) aims to continuously learn new object classes without forgetting previously learned ones. A persistent challenge is catastrophic forgetting, primarily attributed to background shift in conventional detectors. While pseudo-labeling mitigates this in dense detectors, we identify a novel, distinct source of forgetting specific to DETR-like architectures: background foregrounding. This arises from the exhaustiveness constraint of the Hungarian matcher, which forcibly assigns every ground truth target to one prediction, even when predictions primarily cover background regions (i.e., low IoU). This erroneous supervision compels the model to misclassify background features as specific foreground classes, disrupting learned representations and accelerating forgetting. To address this, we propose a Quality-guided Min-Cost Max-Flow (Q-MCMF) matcher. To avoid forced assignments, Q-MCMF builds a flow graph and prunes implausible matches based on geometric quality. It then optimizes for the final matching that minimizes cost and maximizes valid assignments. This strategy eliminates harmful supervision from background foregrounding while maximizing foreground learning signals. Extensive experiments on the COCO dataset under various incremental settings demonstrate that our method consistently outperforms existing state-of-the-art approaches.
Paper Structure (14 sections, 7 equations, 7 figures, 3 tables)

This paper contains 14 sections, 7 equations, 7 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Detection Transformers
  4. Incremental Object Detection
  5. Label Assignment
  6. Method
  7. Preliminary
  8. Quality-guided Min-Cost Max-Flow Matcher
  9. Training Objectives
  10. Experiments
  11. Experimental Setup
  12. Comparison with SOTA Methods
  13. Ablation Study
  14. Conclusion

Figures (7)

  • Figure 1: Quantitative evidence and examples of background foregrounding:(a) shows the proportion of matches with IoU $<$ IoU threshold of Base/New classes at the 40th epoch of 70-10 second phase. (b) shows two examples of background foregrounding.
  • Figure 2: Overall training pipeline of Deformable DETR with Q-MCMF matcher.
  • Figure 3: Effectiveness analysis of Q-MCMF Matcher.
  • Figure 4: Stability-Plasticity Balance with different $\alpha$ and $\beta$.
  • Figure 5: Effectiveness of Q-MCMF Matcher on DN-DETR and DAB-DETR under 70-10 setting.
  • ...and 2 more figures