Research map / Label-efficient and robust learning / Knowledge distillation
GeoSANE: Learning Geospatial Representations from Models, Not Data
Joëlle Hanna, Damian Falk, Stella X. Yu, Damian Borth
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 |
|---|---|---|---|
| initialize with GeoSANE-generated weights instead of random init | ViT-L · fMoW | accuracy: 35.7 -> 58.9 | Improved |
| initialize with GeoSANE-generated weights instead of random init | ViT-L · BigEarthNet | mAP: 69.8 -> 88.7 | Improved |
| initialize with GeoSANE-generated weights instead of random init | Swin-B · DIOR | mAP@0.5: 67.5 -> 79.0 | Improved |
| initialize with GeoSANE-generated weights instead of random init | ViT-L · EuroSAT | accuracy: 95.0 -> 99.1 | Improved |
| use GeoSANE instead of the best RSFM baseline | ViT-L · EuroSAT | accuracy: 99.4 -> 99.1 | Hurt |
| use GeoSANE instead of the best RSFM baseline | Swin-B · DFC2020 | mIoU: 49.8 -> 54.3 | Improved |
| use GeoSANE instead of the best RSFM baseline | Swin-B · Sen1Floods11 | mIoU: 90.9 -> 89.6 | Hurt |
| use learned weight generation instead of DARE merging | GeoSANE · BigEarthNet | mAP: 69.0 -> 88.7 | Improved |
| use GeoSANE-generated lightweight model instead of knowledge distillation | ResNet-18 · BigEarthNet 11M | mAP: 67.3 -> 83.7 | Improved |
| use GeoSANE-generated MobileNetV2 instead of knowledge distillation | MobileNetV2 · fMoW 3.5M | accuracy: 25.5 -> 17.7 | Hurt |
| start from GeoSANE-generated weights instead of the ImageNet prompt | ViT-L · fMoW | accuracy: 52.4 -> 58.9 | Improved |
| start from GeoSANE-generated weights instead of the ImageNet prompt | Swin-B · DIOR | mAP@0.5: 73.6 -> 79.0 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- TerraFM — TerraFM: A Scalable Foundation Model for Unified Multisensor Earth Observation (ICLR 2026) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Towards Faithful Reasoning in Remote Sensing: A Perceptually-Grounded GeoSpatial Chain-of-Thought for Vision-Language Models (ICLR 2026 · similarity 0.88)
- GeoCore-9B: Towards Geo-Aware Generative Foundation Models in Earth Observation (NeurIPS 2026 · similarity 0.88)
- SkyMoE: A Vision-Language Foundation Model for Enhancing Geospatial Interpretation with Mixture of Experts (AAAI 2026 · similarity 0.87)
- GeoLink: Empowering Remote Sensing Foundation Model with OpenStreetMap Data (NeurIPS 2025 · similarity 0.87)
- GeoViS: Geospatially Rewarded Visual Search for Remote Sensing Visual Grounding (CVPR 2026 · similarity 0.87)
- SkySense: A Multi-Modal Remote Sensing Foundation Model Towards Universal Interpretation for Earth Observation Imagery (CVPR 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.
