Research map / Continual learning / Class-incremental
Parameterized Prompt for Incremental Object Detection
Zijia An, Boyu Diao, Ruiqi Liu, Libo Huang, Chuanguang Yang, Fei Wang, Zhulin An, Yongjun Xu
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 pseudo labeling to the baseline (ablation) | Deformable-DETR baseline · VOC2007 5+5+5+5 | AP50 (1-20): 67.4 -> 66.8 | Hurt |
| add parameterized prompt structure (ablation) | baseline with pseudo labeling · VOC2007 5+5+5+5 | AP50 (1-20): 66.8 -> 75.1 | Improved |
| add parameterized prompt fusion (ablation) | P2IOD with parameterized prompts · VOC2007 5+5+5+5 | AP50 (1-20): 75.1 -> 75.3 | Improved |
| add sparse loss (ablation) | P2IOD with fusion · VOC2007 5+5+5+5 | AP50 (1-20): 75.3 -> 76.4 | Improved |
| add pseudo labeling (ablation) | P2IOD with prompts, fusion and sparse loss · VOC2007 5+5+5+5 | AP50 (1-20): 71.5 -> 76.4 | Improved |
| add prompt fusion (ablation) | P2IOD · VOC2007 10+10 | AP50 (1-20): 80.13 -> 81.21 | Improved |
| keep only first-task prompts (top-k 1.0) instead of moderate fusion (ablation) | P2IOD fusion · VOC2007 10+10 | AP50 (1-20): 81.21 -> 80.71 | Hurt |
| set sparse loss lambda to 1e-5 instead of 0 (ablation) | P2IOD · VOC2007 5+5+5+5 | AP50 (1-20): 75.3 -> 76.4 | Improved |
| use too large sparse loss lambda 1e-4 instead of 1e-5 (ablation) | P2IOD · VOC2007 5+5+5+5 | AP50 (1-20): 76.4 -> 76.1 | Hurt |
| replace query-localization with SAC compression (ablation) | MD-DETR · VOC2007 5+5+5+5 | AP50 (1-20): 61.5 -> 64.0 | Improved |
| use average compression instead of SAC (ablation) | P2IOD · VOC2007 5+5+5+5 | AP50 (1-20): 64.7 -> 76.4 | Improved |
| replace MD-DETR prompt pool with parameterized prompts | Deformable-DETR COCO · VOC2007 10+10 | AP50 (1-20): 73.2 -> 81.2 | Improved |
Closest papers by meaning
- Beyond Prompt Degradation: Prototype-guided Dual-pool Prompting for Incremental Object Detection (CVPR 2026 · similarity 0.93)
- Boosting Vision-Language Models Towards Cross-Domain Incremental Object Detection (CVPR 2026 · similarity 0.89)
- Gradient Decomposition and Alignment for Incremental Object Detection (ICCV 2025 · similarity 0.89)
- Incremental Object Detection via Meta-Learning (TPAMI 2021 · similarity 0.88)
- Incremental Learning of Object Detectors Without Catastrophic Forgetting (ICCV 2017 · similarity 0.88)
- Evolving Parameterized Prompt Memory for Continual Learning (AAAI 2024 · similarity 0.87)
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.
