Research map / Detection and segmentation / Open-set and few-shot detection
ORSATR-X: A Foundation Model based on Differential-and-Excitation Networks for Optical Remote Sensing Object Recognition
Canyu Mo, Yongxiang Liu, Jiehua Zhang, Zilong Yu, Zhen Liu, Tianpeng Liu, Li Liu
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 WLA (ablation) | DINOv3-B baseline · DIOR-R | mAP50: 74.64 -> 74.97 | Improved |
| add MSAM (ablation) | DINOv3-B baseline · DIOR-R | mAP50: 74.64 -> 74.81 | Improved |
| combine WLA and MSAM (ablation) | DINOv3-B baseline · DIOR-R | mAP50: 74.64 -> 75.05 | Improved |
| add distillation pre-training on top of WLA+MSAM (ablation) | ORSATR-X · DIOR-R | mAP50: 75.05 -> 75.30 | Improved |
| add WLA and MSAM (ablation) | DINOv3-B baseline · DIOR-R small objects | mAP50: 21.24 -> 23.71 | Improved |
| layer-wise vs block-wise adapter placement (ablation) | DINOv3-B baseline · DIOR-R | mAP50: 74.42 -> 74.64 | Improved |
| WLA instead of standard convolution adapter (ablation) | ORSATR-X · DIOR-R | mAP50: 74.85 -> 74.97 | Improved |
| add WLA over standard conv on medium objects (ablation) | adapter · DIOR-R medium | mAP50: 51.67 -> 51.68 | No clear effect |
| ORSATR-X vs fine-tuned baseline | DINOv3-B · Potsdam | mF1: 91.13 -> 91.36 | Improved |
| ORSATR-X vs fine-tuned baseline | DINOv3-B · DIOR | mAP50: 79.70 -> 80.20 | Improved |
Closest papers by meaning
- InstructSAM: A Training-free Framework for Instruction-Oriented Remote Sensing Object Recognition (NeurIPS 2025 · similarity 0.86)
- SkySense: A Multi-Modal Remote Sensing Foundation Model Towards Universal Interpretation for Earth Observation Imagery (CVPR 2024 · similarity 0.86)
- Learning De-Biased Representations for Remote-Sensing Imagery (NeurIPS 2024 · similarity 0.85)
- ZoomEarth: Active Perception for Ultra-High-Resolution Geospatial Vision-Language Tasks (CVPR 2026 · similarity 0.85)
- AnySat: One Earth Observation Model for Many Resolutions, Scales, and Modalities (CVPR 2025 · similarity 0.84)
- RoMA: Scaling up Mamba-based Foundation Models for Remote Sensing (NeurIPS 2025 · 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.
