Research map / Detection and segmentation / Weakly supervised detection and segmentation
SOAR: Semi-Supervised Open-Vocabulary Aerial Object Detection via Dual-Aware Enhanced Prior Denoising
Xu Liu, Yihong Huang, Dan Zhang, Lingling Li, Long Sun, Licheng Jiao
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 language-aware query enhancer (LAQE) (ablation) | SOAR w/o PAQS · DIOR GZSD | mAP: 45.7 -> 67.7 | Improved |
| add prior-aware query selector (PAQS) alone (ablation) | SOAR · DIOR GZSD | mAP: 45.7 -> 48.4 | Improved |
| add PAQS on top of LAQE (ablation) | SOAR · DIOR GZSD | mAP: 67.7 -> 68.5 | Improved |
| halve queries from 900 to 450 (ablation) | SOAR · DIOR GZSD | mAP: 68.5 -> 64.2, 3.21 -> 5.45 images/second | Mixed |
| initialize background tokens by regional average (GAParea) instead of random (ablation) | SOAR · DIOR GZSD | mAP-novel: 42.2 -> 43.9 | Improved |
| add region positional encodings to background tokens (ablation) | SOAR · DIOR GZSD | mAP-novel: 43.9 -> 44.9 | Improved |
| add multi-granular label extension (average of original and extended labels) (ablation) | SOAR · DIOR GZSD | mAP-novel: 42.3 -> 44.9 | Improved |
| max pooling instead of averaging for label aggregation (ablation) | SOAR · DIOR GZSD | mAP-novel: 44.9 -> 43.8 | Hurt |
| replace CastDet with SOAR on DIOR | CastDet · DIOR GZSD | mAP / HM: 61.6 -> 68.5 mAP; 53.6 -> 55.9 HM | Improved |
| replace CastDet with SOAR on DIOR novel classes | CastDet · DIOR GZSD | mAP-novel: 45.2 -> 44.9 | Hurt |
| replace CastDet with SOAR on DOTA | CastDet · DOTA GZSD | mAP-novel: 29.2 -> 29.5 | No clear effect |
Closest papers by meaning
- Zero-Shot Aerial Object Detection with Visual Description Regularization (AAAI 2024 · similarity 0.87)
- Semi-supervised Open-World Object Detection (AAAI 2024 · similarity 0.87)
- Locate Anything on Earth: Advancing Open-Vocabulary Object Detection for Remote Sensing Community (AAAI 2025 · similarity 0.86)
- Progressive Exploration-Conformal Learning for Sparsely Annotated Object Detection in Aerial Images (NeurIPS 2024 · similarity 0.86)
- Sparse Semi-DETR: Sparse Learnable Queries for Semi-Supervised Object Detection (CVPR 2024 · similarity 0.86)
- InstructSAM: A Training-free Framework for Instruction-Oriented Remote Sensing Object Recognition (NeurIPS 2025 · similarity 0.85)
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.
