AI Paper Atlas

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

CVPR 2026 · Label-efficient and robust learning › Knowledge distillation · approach: Few-shot and metric distillation · 2 citations (OpenAlex)

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 changedWhereMeasuredEffect
initialize with GeoSANE-generated weights instead of random initViT-L · fMoWaccuracy: 35.7 -> 58.9Improved
initialize with GeoSANE-generated weights instead of random initViT-L · BigEarthNetmAP: 69.8 -> 88.7Improved
initialize with GeoSANE-generated weights instead of random initSwin-B · DIORmAP@0.5: 67.5 -> 79.0Improved
initialize with GeoSANE-generated weights instead of random initViT-L · EuroSATaccuracy: 95.0 -> 99.1Improved
use GeoSANE instead of the best RSFM baselineViT-L · EuroSATaccuracy: 99.4 -> 99.1Hurt
use GeoSANE instead of the best RSFM baselineSwin-B · DFC2020mIoU: 49.8 -> 54.3Improved
use GeoSANE instead of the best RSFM baselineSwin-B · Sen1Floods11mIoU: 90.9 -> 89.6Hurt
use learned weight generation instead of DARE mergingGeoSANE · BigEarthNetmAP: 69.0 -> 88.7Improved
use GeoSANE-generated lightweight model instead of knowledge distillationResNet-18 · BigEarthNet 11MmAP: 67.3 -> 83.7Improved
use GeoSANE-generated MobileNetV2 instead of knowledge distillationMobileNetV2 · fMoW 3.5Maccuracy: 25.5 -> 17.7Hurt
start from GeoSANE-generated weights instead of the ImageNet promptViT-L · fMoWaccuracy: 52.4 -> 58.9Improved
start from GeoSANE-generated weights instead of the ImageNet promptSwin-B · DIORmAP@0.5: 73.6 -> 79.0Improved

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.

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

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy