Research map / Detection and segmentation / Open-set and few-shot detection
Plan, Decouple, Assimilate: Physics-Aware Object Insertion in Remote Sensing Imagery
Yingyan Hou, Xianchi Dong, Chao Ren, Wanxuan Lu, Zihan Wei, Hongfeng Yu, Yixiao Wang, Xian 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 |
|---|---|---|---|
| add ASA (solar + frequency) module (ablation) | OminiControl baseline · FAIR1M+iSAID insertion | FID: 12.05 -> 9.752 | Improved |
| add ASA module (ablation) | OminiControl baseline · FAIR1M+iSAID insertion | whole-image PSNR: 25.22 -> 28.37 | Improved |
| add NATA texture assimilation alone (ablation) | OminiControl baseline · FAIR1M+iSAID insertion | FID: 12.05 -> 12.03 | No clear effect |
| add NATA on top of ASA (ablation) | PDA · FAIR1M+iSAID insertion | FID: 9.752 -> 9.732 | Improved |
| add ASA module (ablation) | OminiControl baseline · iSAID evaluation set | shadow direction error: 8.620 -> 5.250 | Improved |
| train on OminiControl synthetic data | Rotated R-CNN · MAR20-11-FewShot | mAP50: 58.13 -> 55.22 | Hurt |
| use PDA synthetic data | S2ANet · MAR20-11-FewShot | mAP50: 55.67 -> 77.36 | Improved |
| raise real:synthetic ratio from 1:1 to 1:4 | YOLO26 · MAR20-11-FewShot | mAP50: peak 84.06 at 1:4 | Improved |
| raise real:synthetic ratio to 1:5 | detectors · MAR20-11-FewShot | mAP50 | No clear effect |
| add recursive multi-object insertion vs single insertion | PDA · MAR20 insertion region | PSNR: 20.34 -> 19.23 | Hurt |
| add ASA to the pipeline (ablation) | OminiControl baseline · A100 inference | time per image (s): 14.5 -> 17.8 | Hurt |
Closest papers by meaning
- In-Context Learning for Remote Sensing Vision: A Semantic-Aware Rotation-Robust Diffusion Framework (NeurIPS 2026 · similarity 0.86)
- SGP2: Coarse-to-Fine Controllable Multimodal Remote Sensing Image Generation (ECCV 2026 · similarity 0.86)
- AnyBand-Diff: A Unified Remote Sensing Image Generation and Band Repair Framework with Spectral Priors (ICML 2026 · similarity 0.86)
- InstructSAM: A Training-free Framework for Instruction-Oriented Remote Sensing Object Recognition (NeurIPS 2025 · similarity 0.85)
- S2MAE: A Spatial-Spectral Pretraining Foundation Model for Spectral Remote Sensing Data (CVPR 2024 · similarity 0.84)
- GeoCore-9B: Towards Geo-Aware Generative Foundation Models in Earth Observation (NeurIPS 2026 · similarity 0.84)
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.
