Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

YOLO-IOD: Towards Real Time Incremental Object Detection

Shizhou Zhang, Xueqiang Lv, Yinghui Xing, Qirui Wu, Di Xu, Chen Zhao, Yanning Zhang

TL;DR

This paper tackles the challenge of incremental object detection for real-time YOLO detectors by identifying three knowledge conflicts—foreground-background confusion, parameter interference, and misaligned knowledge distillation—and proposing a stage-wise, parameter-efficient framework, YOLO-IOD, built on pretrained YOLO-World. It introduces three modules: Conflict-Aware Pseudo-Label Refinement (CPR), Importance-based Kernel Selection (IKS), and Cross-Stage Asymmetric Knowledge Distillation (CAKD), plus a realistic LoCo COCO benchmark to avoid data leakage. Experimental results on COCO and LoCo COCO demonstrate state-of-the-art performance with minimal forgetting across single-step and multi-step settings, while maintaining real-time inference. The LoCo COCO benchmark provides a more practical evaluation by aligning category co-occurrence and removing inter-stage image overlap, enhancing the field’s ability to assess real-world IOD performance.

Abstract

Current methods for incremental object detection (IOD) primarily rely on Faster R-CNN or DETR series detectors; however, these approaches do not accommodate the real-time YOLO detection frameworks. In this paper, we first identify three primary types of knowledge conflicts that contribute to catastrophic forgetting in YOLO-based incremental detectors: foreground-background confusion, parameter interference, and misaligned knowledge distillation. Subsequently, we introduce YOLO-IOD, a real-time Incremental Object Detection (IOD) framework that is constructed upon the pretrained YOLO-World model, facilitating incremental learning via a stage-wise parameter-efficient fine-tuning process. Specifically, YOLO-IOD encompasses three principal components: 1) Conflict-Aware Pseudo-Label Refinement (CPR), which mitigates the foreground-background confusion by leveraging the confidence levels of pseudo labels and identifying potential objects relevant to future tasks. 2) Importancebased Kernel Selection (IKS), which identifies and updates the pivotal convolution kernels pertinent to the current task during the current learning stage. 3) Cross-Stage Asymmetric Knowledge Distillation (CAKD), which addresses the misaligned knowledge distillation conflict by transmitting the features of the student target detector through the detection heads of both the previous and current teacher detectors, thereby facilitating asymmetric distillation between existing and newly introduced categories. We further introduce LoCo COCO, a more realistic benchmark that eliminates data leakage across stages. Experiments on both conventional and LoCo COCO benchmarks show that YOLO-IOD achieves superior performance with minimal forgetting.

YOLO-IOD: Towards Real Time Incremental Object Detection

TL;DR

This paper tackles the challenge of incremental object detection for real-time YOLO detectors by identifying three knowledge conflicts—foreground-background confusion, parameter interference, and misaligned knowledge distillation—and proposing a stage-wise, parameter-efficient framework, YOLO-IOD, built on pretrained YOLO-World. It introduces three modules: Conflict-Aware Pseudo-Label Refinement (CPR), Importance-based Kernel Selection (IKS), and Cross-Stage Asymmetric Knowledge Distillation (CAKD), plus a realistic LoCo COCO benchmark to avoid data leakage. Experimental results on COCO and LoCo COCO demonstrate state-of-the-art performance with minimal forgetting across single-step and multi-step settings, while maintaining real-time inference. The LoCo COCO benchmark provides a more practical evaluation by aligning category co-occurrence and removing inter-stage image overlap, enhancing the field’s ability to assess real-world IOD performance.

Abstract

Current methods for incremental object detection (IOD) primarily rely on Faster R-CNN or DETR series detectors; however, these approaches do not accommodate the real-time YOLO detection frameworks. In this paper, we first identify three primary types of knowledge conflicts that contribute to catastrophic forgetting in YOLO-based incremental detectors: foreground-background confusion, parameter interference, and misaligned knowledge distillation. Subsequently, we introduce YOLO-IOD, a real-time Incremental Object Detection (IOD) framework that is constructed upon the pretrained YOLO-World model, facilitating incremental learning via a stage-wise parameter-efficient fine-tuning process. Specifically, YOLO-IOD encompasses three principal components: 1) Conflict-Aware Pseudo-Label Refinement (CPR), which mitigates the foreground-background confusion by leveraging the confidence levels of pseudo labels and identifying potential objects relevant to future tasks. 2) Importancebased Kernel Selection (IKS), which identifies and updates the pivotal convolution kernels pertinent to the current task during the current learning stage. 3) Cross-Stage Asymmetric Knowledge Distillation (CAKD), which addresses the misaligned knowledge distillation conflict by transmitting the features of the student target detector through the detection heads of both the previous and current teacher detectors, thereby facilitating asymmetric distillation between existing and newly introduced categories. We further introduce LoCo COCO, a more realistic benchmark that eliminates data leakage across stages. Experiments on both conventional and LoCo COCO benchmarks show that YOLO-IOD achieves superior performance with minimal forgetting.
Paper Structure (22 sections, 6 equations, 3 figures, 4 tables)

This paper contains 22 sections, 6 equations, 3 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Incremental Learning
  4. Incremental Object Detection
  5. Preliminaries and Benchmark
  6. Problem Definition
  7. YOLO-World
  8. LoCo COCO Benchmark
  9. Method
  10. Overall Framework
  11. Conflict-Aware Pseudo-Label Refinement
  12. Importance-based Kernel Selection
  13. Cross-Stage Asymmetric Knowledge Distillation
  14. Experiments
  15. Experiment Setup
  16. ...and 7 more sections

Figures (3)

  • Figure 2: Overall architecture of YOLO-IOD. Our method integrates three components to address knowledge conflicts in IOD: 1) Conflict-aware Pseudo-label Refinement (CPR); 2) Importance-based Kernel Selection (IKS); and 3) Cross-Stage Asymmetric Knowledge Distillation (CAKD).
  • Figure 3: Ablation study within the CAKD module on the COCO 20-20 incremental setting.
  • Figure 4: Ablation of kernel selection ratio $\mathcal{K}$ in IKS on COCO 20-20 incremental setting.