Research map / Detection and segmentation / Open-set and few-shot detection
Exploiting Unlabeled Data with Multiple Expert Teachers for Open Vocabulary Aerial Object Detection and Its Orientation Adaptation
Yanfang Li, Weiwei Guo, Xue Yang, Ning Liao, Shaofeng Zhang, Yi Yuan Yu, Wenxian Yu, Junchi Yan
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 external CLIP teacher (hybrid training) to student-teacher training (ablation) | CastDet · VisDroneZSD | mAR novel: 47.7 -> 69.1 | Improved |
| select pseudo-boxes with regression jittering variance instead of RPN score (ablation) | CastDet · VisDroneZSD | mAP novel: 43.3 -> 47.8 | Improved |
| select pseudo-boxes with box jittering variance instead of RPN score (ablation) | CastDet · VisDroneZSD | mAP novel: 43.3 -> 43.6 | No clear effect |
| combine AJV and SJV scores for oriented box selection (ablation) | CastDet oriented · DOTA | mAP novel: 27.6 -> 34.9 | Improved |
| add dynamic label queue (dynamic update plus queue) (ablation) | CastDet · VisDroneZSD | mAP: 11.6 -> 39.5 | Improved |
| add a queue on top of dynamic update (ablation) | CastDet · VisDroneZSD | mAP novel: 43.5 -> 43.3 | No clear effect |
| train GLIP with our pseudo-labels | GLIP · DOTA | mAP novel: 8.0 -> 14.6 | Improved |
| train GroundingDINO with our pseudo-labels | GroundingDINO · DOTA | mAP: 58.0 -> 47.0 | Mixed |
| use ViT-L-14 RemoteCLIP instead of R50 teacher (ablation) | CastDet · VisDroneZSD | mAP novel: 46.3 -> 51.2 | Improved |
| reduce labeled data to 34% (ablation) | CastDet · VisDroneZSD | mAP: 39.5 -> 38.6 | No clear effect |
| raise loss weight beta to 2.0 (ablation) | CastDet · DOTA | mAP novel: 29.6 -> 34.9 | Mixed |
| use prompt T4 instead of plain category name (ablation) | CastDet · DOTA | mAP novel: 24.0 -> 34.9 | Improved |
Closest papers by meaning
- Toward Open Vocabulary Aerial Object Detection with CLIP-Activated Student-Teacher Learning (ECCV 2024 · similarity 0.90)
- SegEarth-OV: Towards Training-Free Open-Vocabulary Segmentation for Remote Sensing Images (CVPR 2025 · similarity 0.84)
- FRED: Towards a Full Rotation-Equivariance in Aerial Image Object Detection (AAAI 2024 · similarity 0.83)
- Progressive Exploration-Conformal Learning for Sparsely Annotated Object Detection in Aerial Images (NeurIPS 2024 · similarity 0.82)
- DitHub: A Modular Framework for Incremental Open-Vocabulary Object Detection (NeurIPS 2025 · similarity 0.82)
- Locate Anything on Earth: Advancing Open-Vocabulary Object Detection for Remote Sensing Community (AAAI 2025 · similarity 0.81)
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.
