Versatile Vision Transformer that uses a single Universal Patch Encoder to process satellite data from any sensor, modality, or resolution.</p>\n","updatedAt":"2026-06-23T04:56:48.399Z","author":{"_id":"662b7fba68ed7bbf40bfb0df","avatarUrl":"/avatars/c04f8fd10bc78a6613d5a0c74bda3cef.svg","fullname":"Guillaume Astruc","name":"g-astruc","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":5,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.8349064588546753},"editors":["g-astruc"],"editorAvatarUrls":["/avatars/c04f8fd10bc78a6613d5a0c74bda3cef.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2606.23503","authors":[{"_id":"6a3a0824fdcd3514343bb5ec","name":"Yohann Perron","hidden":false},{"_id":"6a3a0824fdcd3514343bb5ed","user":{"_id":"662b7fba68ed7bbf40bfb0df","avatarUrl":"/avatars/c04f8fd10bc78a6613d5a0c74bda3cef.svg","isPro":false,"fullname":"Guillaume Astruc","user":"g-astruc","type":"user","name":"g-astruc"},"name":"Guillaume Astruc","status":"claimed_verified","statusLastChangedAt":"2026-06-23T13:56:20.479Z","hidden":false},{"_id":"6a3a0824fdcd3514343bb5ee","name":"Nicolas Gonthier","hidden":false},{"_id":"6a3a0824fdcd3514343bb5ef","name":"Clement Mallet","hidden":false},{"_id":"6a3a0824fdcd3514343bb5f0","name":"Loic Landrieu","hidden":false}],"publishedAt":"2026-06-22T00:00:00.000Z","submittedOnDailyAt":"2026-06-23T00:00:00.000Z","title":"UniverSat: Resolution- and Modality-Agnostic Transformers for Earth Observation","submittedOnDailyBy":{"_id":"662b7fba68ed7bbf40bfb0df","avatarUrl":"/avatars/c04f8fd10bc78a6613d5a0c74bda3cef.svg","isPro":false,"fullname":"Guillaume Astruc","user":"g-astruc","type":"user","name":"g-astruc"},"summary":"Vision Transformers (ViT) dominate computer vision. However, their reliance on rigid patch projectors hinders transfer to Earth Observation (EO), where input modalities, scales, and resolutions vary widely. We introduce UniverSat, a ViT-style backbone built around a Universal Patch Encoder that maps patches from arbitrary spatial, spectral, and temporal resolutions, and from both optical and non-optical sensors, into a shared embedding space with a shared set of weights. This enables training a single model on heterogeneous multimodal corpora via self-supervision, yielding robust, sensor-agnostic spatial features. We validate this approach with strong results across classification and segmentation on standard EO benchmarks from GeoBench, PANGEABench, and SpectralEarth. Our code and models are available at https://github.com/gastruc/UniverSat.","upvotes":2,"discussionId":"6a3a0825fdcd3514343bb5f1","projectPage":"https://gastruc.github.io/universat","githubRepo":"https://github.com/gastruc/UniverSat","githubRepoAddedBy":"user","ai_summary":"UniverSat introduces a Universal Patch Encoder for Vision Transformers that enables robust, sensor-agnostic spatial feature extraction across diverse Earth Observation data types.","ai_keywords":["Vision Transformers","Universal Patch Encoder","patch projectors","Earth Observation","multimodal corpora","self-supervision","spatial features","classification","segmentation"],"ai_summary_model":"Qwen/Qwen2.5-Coder-32B-Instruct","githubStars":20},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"662b7fba68ed7bbf40bfb0df","avatarUrl":"/avatars/c04f8fd10bc78a6613d5a0c74bda3cef.svg","isPro":false,"fullname":"Guillaume Astruc","user":"g-astruc","type":"user"},{"_id":"654bb2591a9e65ef2598d8c4","avatarUrl":"/avatars/b444ac6ac3280bfad1022af0c404dad1.svg","isPro":false,"fullname":"osv5m","user":"osv5m","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"query":{}}">

Cite arxiv.org/abs/2606.23503 in a dataset README.md to link it from this page.

Cite arxiv.org/abs/2606.23503 in a Space README.md to link it from this page.

Add this paper to a collection to link it from this page.