Research map / Continual learning / Class-incremental
Gradient Decomposition and Alignment for Incremental Object Detection
Wenlong Luo, Shizhou Zhang, De Cheng, Yinghui Xing, Guoqiang Liang, Peng Wang, Yanning Zhang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add GM-Pseudo Gaussian mixture pseudo-labeling (ablation) | pseudo-labeling baseline · VOC2007 5-5 | mAP classes 1-20: 58.5 -> 61.7 | Improved |
| add GM-Pseudo (ablation) | pseudo-labeling baseline · VOC2007 10-5 | mAP classes 1-20: 68.1 -> 68.6 | Improved |
| add gradient decomposition and alignment (GDA) (ablation) | GM-Pseudo model · VOC2007 5-5 | mAP classes 1-20: 61.7 -> 63.5 | Improved |
| add GDA (ablation) | GM-Pseudo model · VOC2007 10-5 | mAP classes 1-20: 68.6 -> 69.3 | Improved |
| add GDA (ablation) | GM-Pseudo model · VOC2007 5-5 | mAP classes 1-5: 61.0 -> 62.2 | Improved |
| set lambda to 0.6 rather than larger values (ablation) | GDA · COCO 70+10 | AP | Improved |
| set alpha to 1.0 (VOC) or 2.0 (COCO) in GM-Pseudo (ablation) | GM-Pseudo · VOC 5-5 / COCO 70-10 | mAP / AP | Improved |
| use GDA-IOD without exemplar replay versus replay-based ABR | GDA-IOD · VOC2007 5+5+5+5 (4 tasks) | mAP classes 6-20: over 7% mAP | Improved |
| compare GDA-IOD to BPF on old classes | GDA-IOD · VOC2007 10+10 (single step) | mAP old classes: 3.4% | Improved |
| compare GDA-IOD to BPF on 19+1 old classes | GDA-IOD · VOC2007 19+1 | mAP new class: 65.3 -> 58.4 | Hurt |
Closest papers by meaning
- YOLO-IOD: Towards Real Time Incremental Object Detection (AAAI 2026 · similarity 0.92)
- Incremental Object Detection via Future-Aware Decoupled Cross-Head Distillation (CVPR 2026 · similarity 0.92)
- Demystifying Catastrophic Forgetting in Two-Stage Incremental Object Detector (ICML 2025 · similarity 0.89)
- Revisiting Generative Replay for Class Incremental Object Detection (CVPR 2025 · similarity 0.89)
- GCD: Advancing Vision-Language Models for Incremental Object Detection via Global Alignment and Correspondence Distillation (AAAI 2025 · similarity 0.89)
- DCA: Dividing and Conquering Amnesia in Incremental Object Detection (AAAI 2025 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
