Research map / Detection and segmentation / Open-set and few-shot detection
VK-Det: Visual Knowledge Guided Prototype Learning for Open-Vocabulary Aerial Object Detection
Jianhang Yao, Yongbin Zheng, Siqi Lu, Wanying Xu, Peng Sun
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 |
|---|---|---|---|
| VK-Det without extra supervision vs CastDet with extra supervision | open-vocab aerial detector · DIOR (16 base/4 novel) | mAP_N (novel): 29.8 -> 30.1 | Improved |
| VK-Det vs CastDet | open-vocab aerial detector · DOTA (11 base/4 novel) | mAP_N (novel): 14.2 -> 23.3 | Improved |
| add PAPL prototype pseudo-labeling to ASKD (ablation) | VK-Det · DIOR (16 base/4 novel) | mAP_N (novel): 7.8 -> 20.4 | Improved |
| add SMI localization scoring to ASKD (ablation) | VK-Det · DIOR (16 base/4 novel) | mAP_N (novel): 7.8 -> 20.1 | Improved |
| combine ASKD, PAPL and SMI (full) (ablation) | VK-Det · DIOR (16 base/4 novel) | mAP_N (novel): 20.4 -> 30.1 | Improved |
| add max-min edge jitter data augmenter (ablation) | ASKD · DIOR (16 base/4 novel) | mAP_N (novel): 20.0 -> 23.2 | Improved |
| add attention-mask proposal selection (ablation) | ASKD · DIOR (16 base/4 novel) | mAP_N (novel): 20.0 -> 24.5 | Improved |
| use prototype pseudo-labels instead of text-embedding (extra-supervision) pseudo-labels (ablation) | PAPL · DIOR (16 base/4 novel) | mAP_N (novel): 28.1 -> 30.1 | Improved |
| add distillation score to prototype/localization scores in SMI (ablation) | SMI · DIOR (16 base/4 novel) | mAP_N (novel): 12.6 -> 30.1 | Improved |
| add prototype score to distillation/localization scores in SMI (ablation) | SMI · DIOR (16 base/4 novel) | mAP_N (novel): 24.8 -> 30.1 | Improved |
| add PAPL / SMI to ASKD (ablation) | VK-Det · DIOR (16 base/4 novel) | mAP_B (base): 69.9 -> 64.4 | Hurt |
Closest papers by meaning
- Cyclic Contrastive Knowledge Transfer for Open-Vocabulary Object Detection (ICLR 2025 · similarity 0.89)
- Zero-Shot Aerial Object Detection with Visual Description Regularization (AAAI 2024 · similarity 0.88)
- Open-Det: An Efficient Learning Framework for Open-Ended Detection (ICML 2025 · similarity 0.88)
- Generative Region-Language Pretraining for Open-Ended Object Detection (CVPR 2024 · similarity 0.87)
- Simple Image-Level Classification Improves Open-Vocabulary Object Detection (AAAI 2024 · similarity 0.87)
- LLMs Meet VLMs: Boost Open Vocabulary Object Detection with Fine-grained Descriptors (ICLR 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.
