Research map / Continual learning / Class-incremental
Orthogonal Knowledge Refreshing for Domain-Incremental Object Detection
Aoting Zhang, Dongbao Yang, Chang Liu, Xiaopeng Hong, Can Ma, Yu ZHOU
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 backbone fine-tuning (ablation) | Baseline detector · Pascal VOC series Session 4 | mAP@0.5: 44.2 -> 49.5 | Improved |
| Add low-rank subspace expansion (SE) (ablation) | FT-Back · Pascal VOC series Session 4 | mAP@0.5: 49.5 -> 60.9 | Improved |
| Add gradient-based orthogonal refreshing (GOR) (ablation) | FT-Back + SE · Pascal VOC series Session 4 | mAP@0.5: 60.9 -> 64.9 | Improved |
| Add topology-aware consistency (TAC) (ablation) | FT-Back + SE · Pascal VOC series Session 4 | mAP@0.5: 60.9 -> 63.3 | Improved |
| Add TAC without SE (ablation) | FT-Back · Pascal VOC series Session 4 | mAP@0.5: 49.5 -> 60.2 | Improved |
| Add TAC on top of SE and GOR (ablation) | FT-Back + SE + GOR · Pascal VOC series Session 4 | mAP@0.5: 64.9 -> 65.2 | Improved |
| Add GOR on top of SE and TAC (ablation) | FT-Back + SE + TAC · Pascal VOC series Session 4 | mAP@0.5: 63.3 -> 65.2 | Improved |
| Use smaller LoRA ranks (8 instead of 16) (ablation) | OKR · Pascal VOC series | mAP@0.5 | Hurt |
| Increase LoRA rank beyond 16 (to 64) (ablation) | OKR · Pascal VOC series | mAP@0.5 | No clear effect |
| Add orthogonal projection (GOR) to training (ablation) | OKR · Training loss convergence | training loss | No clear effect |
| Use OKR instead of LDB+SOYO | Exemplar-free DIOD · Pascal VOC series Session 4 | mAP@0.5: 59.6 -> 65.2 | Improved |
| Use OKR instead of LDB+SOYO | Exemplar-free DIOD · BDD100K series Session 3 | mAP@0.5: 51.7 -> 58.2 | Improved |
Closest papers by meaning
- Non-exemplar Domain Incremental Object Detection via Learning Domain Bias (AAAI 2024 · similarity 0.91)
- SCAN: Cross Domain Object Detection with Semantic Conditioned Adaptation (AAAI 2022 · similarity 0.90)
- SEEN-DA: SEmantic ENtropy guided Domain-aware Attention for Domain Adaptive Object Detection (CVPR 2025 · similarity 0.89)
- DA-Ada: Learning Domain-Aware Adapter for Domain Adaptive Object Detection (NeurIPS 2024 · similarity 0.88)
- Universal Domain Adaptive Object Detection via Dual Probabilistic Alignment (AAAI 2025 · similarity 0.88)
- GCD: Advancing Vision-Language Models for Incremental Object Detection via Global Alignment and Correspondence Distillation (AAAI 2025 · similarity 0.88)
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.
