Research map / Detection and segmentation / Open-set and few-shot detection
Parameter-Efficient Semantic Augmentation for Enhancing Open-Vocabulary Object Detection
Weihao Cao, Runqi Wang, Xiaoyue Duan, Jinchao Zhang, Ang Yang, Liping Jing
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 V-LoRA to zero-shot model (ablation) | OV-DINO · ArTaxOr + OV-COCO (PEFT) | mAP target: 1.4 -> 61.6 | Improved |
| add V-LoRA to zero-shot model (ablation) | OV-DINO · ArTaxOr + OV-COCO (PEFT) | mAP OV-COCO: 50.6 -> 22.7 | Hurt |
| add MSPB on top of V-LoRA (ablation) | HSA-DINO · ArTaxOr + OV-COCO (PEFT) | mAP target: 61.6 -> 79.1 | Improved |
| add MSPB on top of V-LoRA without SAR (ablation) | HSA-DINO · ArTaxOr + OV-COCO (PEFT) | mAP OV-COCO: 22.7 -> 0.5 | Hurt |
| add Semantic-Aware Router (ablation) | HSA-DINO (V-LoRA+MSPB) · ArTaxOr + OV-COCO (PEFT) | H: 1.0 -> 60.5 | Improved |
| use MSPB instead of CoOp prompts (ablation) | HSA-DINO · Vertical domains mean H | Hmean: 52.1 -> 54.4 | Improved |
| use MSPB instead of AttriCLIP (ablation) | HSA-DINO · Vertical domains mean H | Hmean: 53.0 -> 54.4 | Improved |
| use MSPB instead of predefined prompt (ablation) | HSA-DINO · Vertical domains mean H | Hmean: 49.9 -> 54.4 | Improved |
| use SAR instead of DDAS router (ablation) | HSA-DINO · Vertical domains mean H | Hmean: 46.2 -> 54.4 | Improved |
| add orthogonal prompt loss Lp (ablation) | HSA-DINO · Mean mAP over 3 downstream datasets | mAP avg: 61.5 -> 61.9 | Improved |
| use bank size N=10 (ablation) | MSPB · downstream datasets | mAP avg | Improved |
| use HSA-DINO instead of OV-DINO LoRA | OV-DINO · ArTaxOr + OV-COCO | H: 36.8 -> 60.5 | Improved |
Closest papers by meaning
- CAKE: Category Aware Knowledge Extraction for Open-Vocabulary Object Detection (AAAI 2025 · similarity 0.93)
- SHiNe: Semantic Hierarchy Nexus for Open-vocabulary Object Detection (CVPR 2024 · similarity 0.92)
- DitHub: A Modular Framework for Incremental Open-Vocabulary Object Detection (NeurIPS 2025 · similarity 0.91)
- Open-Det: An Efficient Learning Framework for Open-Ended Detection (ICML 2025 · similarity 0.91)
- Simple Image-Level Classification Improves Open-Vocabulary Object Detection (AAAI 2024 · similarity 0.91)
- Cyclic Contrastive Knowledge Transfer for Open-Vocabulary Object Detection (ICLR 2025 · similarity 0.91)
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.
