Our Datasets

Year	Dataset / Benchmark / Product	Assosiated publication	Type	Scale / key metric	Data/Resource
2026	GlobalBuildingMap (GBM)	GlobalBuildingMap—Unveiling the Mystery of Global Buildings	Global building product	3 m global raster; 790,101 PlanetScope images; >100k training pairs from 74 cities	Paper/Data
2026	GeoHeight-Bench / GeoHeight-Bench+	GeoHeight-Bench: Towards Height-Aware Multimodal Reasoning in Remote Sensing	2 height-reasoning benchmarks	2 complementary benchmarks: relative-height + holistic terrain-aware reasoning; VLM-driven data generation	Paper
2026	OceanTACO	OceanTACO: A Multi-Sensor Global Ocean Sea Surface State Dataset	Global multisensor ocean dataset	Global harmonized ocean-surface observations; core Mar 2023–Aug 2025; extended record to Jan 2015	Paper
2026	GlobalGeoTree	GlobalGeoTree: A Multi-Granular Vision-Language Dataset for Global Tree Species Classification	Biodiversity vision-language dataset	6.3M geolocated tree occurrences; 275 families; 2,734 genera; 21,001 species; S2 time series + 27 environmental variables	Dataset
2026	GAIA	GAIA: A Global, Multimodal, Multiscale Vision-Language Dataset for Remote Sensing Image Analysis	Vision-language dataset	205,150 image-text pairs; global 25-year coverage; 5 scientific captions/image; multimodal + multiscale	Paper
2026	CropClimateX / Climate-Aware US Crop Dataset	A Large-Scale, Multitask, Multisensory Dataset for Climate-Aware Crop Monitoring in the US from 2018–2022	Crop/climate dataset	15,500 12×12 km data cubes; 1,527 US counties; 2018–2022; >6.3 TB; multisensor + climate/soil/terrain variables	Dataset
2026	TUM2TWIN	TUM2TWIN: Introducing the Large-Scale Multimodal Urban Digital Twin Benchmark Dataset	Urban digital-twin benchmark	32 multimodal subsets; ~100,000 m²; 767 GB; terrestrial/mobile/aerial/satellite + aligned 3D data	Paper
2026	EuroMineNet	EuroMineNet: A Multitemporal Sentinel-2 Benchmark for Spatiotemporal Mining Footprint Analysis in the EU (2015–2024)	Multitemporal mining benchmark	133 mining sites; annual S2 + expert labels over 10 years; 2 benchmark tasks; 20 DL models	Paper
2026	GEO-Bench-2	GEO-Bench-2: From Performance to Capability, Rethinking Evaluation in Geospatial AI	Benchmark suite	19 permissively licensed datasets; classification, segmentation, regression, object detection, instance segmentation	Paper
2026	Copernicus-Embed-025deg	ISPRS Congress 2026 contribution	Global EO embedding product	Global 0.25° grid; 721×1440×768 embeddings	Repository
2026	SpectralEarth-MM	SpectralEarth-FM: Bringing Hyperspectral Imagery into Multimodal Earth Observation Pretraining	Global multimodal pretraining dataset	~2M locations; ~25M patches; >40 TB; hyperspectral + optical + SAR + thermal	Paper
2025/26	China Major Cities Urban Green Space Dataset	Enhancing the Understanding of Urban Green Space Cooling Effects in China’s Major Cities with a Sub-Meter Dataset	National urban-green product	49 major Chinese cities; sub-meter resolution	Paper
2025	REO-Instruct	Towards Unified Vision Language Models for Forest Ecological Analysis in Earth Observation	EO instruction/V-L dataset	~1.6M train + ~20k validation + ~36k test image-text pairs; RGB + S2 + ALOS-2 SAR; 4 linked tasks	Repository
2025	ImpactMesh	Technical report forthcoming	Multimodal disaster benchmark	435 flood & wildfire events globally; 41,090 samples; 4 temporal observations/event; S1 + S2 + DEM + Copernicus EMS annotations	Dataset
2025	ChatEarthBench	ChatEarthBench: Benchmarking Multimodal Large Language Models for Earth Observation	MLLM benchmark suite	10 EO image-text datasets across 3 modalities	Repository
2025	Global Vegetation Water Content	Deep Learning-Based GNSS-R Global Vegetation Water Content: Dataset, Estimation, and Uncertainty	Global GNSS-R data product	CYGNSS–GLDAS–SMAP triplets over >3 years; 12-month independent test; daily global VWC product	Paper
2025	CGS	CGS: A Triplet Dataset for GNSS-R Vegetation Monitoring	GNSS-R benchmark dataset	CYGNSS + GLDAS + SMAP triplets; >3-year temporal coverage	[VERIFY standalone data page]
2025	Sen12Landslides	A Spatio-Temporal Dataset for Satellite-Based Landslide Detection	Landslide benchmark	~75k annotations; 15 global regions; 12k+ patches; S1 + S2 + DEM	Paper/Data
2025	GlobalBuildingAtlas (GBA)	GlobalBuildingAtlas: An Open Global and Complete Dataset of Building Polygons, Heights and LoD1 3D Models	Global 2D/3D building product	2.75B+ building polygons; heights + LoD1 3D models; ~97% coverage	Paper
2025	FloodCastBench	FloodCastBench: A Large-Scale Dataset and Foundation Models for Flood Modeling and Forecasting	Flood benchmark	4 major floods; 30/60/480 m products; up to 30 m spatial / 300 s temporal resolution	Repository
2025	ChatEarthNet	ChatEarthNet: A Global-Scale Image-Text Dataset Empowering Vision-Language Geo-Foundation Models	EO image-text dataset	163,488 GPT-3.5 pairs + 10,000 GPT-4V pairs	[VERIFY dataset page]
2025	MGRS-200k	FarSLIP: Discovering Effective CLIP Adaptation for Fine-Grained Remote Sensing Understanding	Fine-grained V-L dataset	~200k scale; multi-granularity captions + object-level textual supervision; ~105 GB	Paper
2025	GeoLangBind-2M	GeoLangBind: Unifying Earth Observation with Agglomerative Vision-Language Foundation Models	EO vision-language dataset	~2M image-text pairs; 6 EO modalities	Repository
2025	ExEBench	ExEBench: Benchmarking Foundation Models on Extreme Earth Events	Extreme-event benchmark suite	7 extreme-event categories	Paper
2025	REOBench	REOBench: Benchmarking Robustness of Earth Observation Foundation Models	Robustness benchmark	6 EO tasks × 12 corruption types	Repository
2025	Copernicus-Bench	Towards a Unified Copernicus Foundation Model for Earth Vision	Multimodal benchmark suite	15 downstream datasets/tasks; 6 newly curated datasets; S1/S2/S3/S5P	Dataset
2025	Copernicus-Pretrain / SSL4EO-S	same	Global multimodal pretraining dataset	18.7M aligned images; ~310k 0.25° global grids; S1/S2/S3/S5P + DEM	Dataset
2025	SpectralEarth	SpectralEarth: Training Hyperspectral Foundation Models at Scale	Hyperspectral pretraining + benchmarks	538,974 patches; 415,153 locations; 11,636 EnMAP scenes; >3 TB; 9 downstream benchmarks	Project
2024	QuickQuakeBuildings (QQB)	QuickQuakeBuildings: Post-earthquake SAR-Optical Dataset for Quick Damaged-building Detection	Disaster SAR–optical benchmark	4,000+ buildings; co-registered VHR SAR + optical + damage labels	Repository
2024/25	EuroSAT-SAR	Feature Guided Masked Autoencoder for Self-Supervised Learning in Remote Sensing	SAR benchmark	27,000 Sentinel-1 images; 10 balanced classes; geographically matched to EuroSAT	Dataset
2024	UTCSA	Continent-wide Urban Tree Canopy Fine-scale Mapping and Coverage Assessment in South America…	Continental tree-canopy product	0.5 m; 888 South American cities	[VERIFY data page]
2024	So2Sat-BuildingType	associated So2Sat building-type work	EO + social-media + OSM dataset	6.95M OSM-labeled buildings; 26.67M geotagged tweets; 42 cities	Repository
2024	Svalbard Calving Front Dataset	A High-Resolution Calving Front Data Product for Marine-Terminating Glaciers in Svalbard	Cryosphere data product	124,919 calving-front traces; 149 glaciers; 1985–2023; Landsat/ASTER/S1/S2	Dataset
2024	RegressionUQ	How Certain Are Uncertainty Estimates? Three Novel Earth Observation Datasets…	UQ regression benchmark	40,000 simulated biomass samples with known aleatoric uncertainty	Paper
2024	SegmentationUQ	same	UQ segmentation benchmark	10,000 base Berlin patches; 3 noise types × 4 levels × 50 realizations; ~6M noisy patches	Paper
2024	ClassificationUQ	same	UQ classification benchmark	10 European cities; 10 independent expert votes per image; probabilistic LCZ labels	Paper
2024	MineNetCD	MineNetCD: A Benchmark for Global Mining Change Detection on Remote Sensing Imagery	Change-detection benchmark	>70k bi-temporal HR pairs; 100 mining sites; pixel-wise labels; 13+ CD models	Paper
2024	UrbanSARFloods	UrbanSARFloods…	SAR flood benchmark	8,879 512×512 chips; 807,500 km²; 18 floods; 5 continents; 20 land-cover classes	Paper
2024	RefSegRS	RRSIS: Referring Remote Sensing Image Segmentation	V-L segmentation dataset	4,420 image-expression-mask triplets; 285 scenes; 14 classes; 0.13 m	Dataset info
2023	Five-Billion-Pixels (FBP)	Enabling Country-Scale Land Cover Mapping with Meter-Resolution Satellite Imagery	Large-scale HR land-cover dataset	>5B manually labeled pixels; 150 GF-2 images; 4 m resolution; >50,000 km²; 24 classes; 60+ administrative districts	Project/Data
2023	UTB – Urban Tree Canopy Brazil	Nationwide Urban Tree Canopy Mapping and Coverage Assessment in Brazil…	National tree-canopy product	0.5 m; 472 Brazilian cities	Project
2023	MedSat	MedSat: A Public Health Dataset for England Featuring Medical Prescriptions and Satellite Imagery	EO + public-health dataset	England LSOAs; 2019–2020; 111 sociodemographic + 43 environmental variables; seasonal S2 composites; prescription variables	Dataset
2023	C2Seg / Cross-City Benchmark	Cross-City Matters…	Cross-city multimodal segmentation benchmark	Berlin–Augsburg + Beijing–Wuhan; HSI/MSI/SAR; cross-city domain-adaptation setup	Paper
2023	MDAS	MDAS: A New Multimodal Benchmark Dataset for Remote Sensing	Multimodal benchmark	5 modalities; 121.7 km² Augsburg; SAR/MSI/HSI/DSM/GIS; 3 benchmark applications	Dataset
2023	RSSOD-Bench	RSSOD-Bench…	Salient-object benchmark	4 US cities; pixel-level labels; 23 methods benchmarked	Paper
2023	GAMUS	GAMUS…	Geometry-aware multimodal benchmark	11,507 RGB+nDSM tiles; 5 cities; 0.33 m	Paper
2023	GEO-Bench	GEO-Bench: Toward Foundation Models for Earth Monitoring	EO-FM benchmark suite	12 tasks: 6 classification + 6 segmentation; 20 baselines	Paper
2023	SSL4EO-L	SSL4EO-L: Datasets and Foundation Models for Landsat Imagery	Pretraining + benchmark family	5M Landsat patches; 3 sensors; 2 product levels; multiple downstream benchmark datasets	Paper
2023	SSL4EO-S12	SSL4EO-S12…	Global multimodal/multitemporal pretraining dataset	251,079 locations × 4 seasons; S1 + S2 L1C/L2A; ~1.5 TB	Repository
2022	RSMSS / Parse Semantics from Geometry	Parse Semantics from Geometry…	RGB-height segmentation benchmark	9,340 RGB+nDSM tiles; 3 US cities; 1024×1024; 6 classes	Dataset
2022	MultiScene / MultiScene-Clean	MultiScene…	Multi-label aerial benchmark	100,000 aerial images; 36 classes; 0.3–0.6 m/px; 14,000 manually verified MultiScene-Clean subset	Project
2022	ReforesTree	ReforesTree…	Forest/carbon dataset	6 Ecuadorian sites; 100 RGB drone images at 2 cm/px; >4,600 tree crowns with species/DBH/AGB/carbon labels	Repository
2022	So2Sat GUL	The Urban Morphology on Our Planet – Global Perspectives from Space	Global urban morphology/LCZ product	LCZ maps + analysis-ready S1/S2 for 1,692 cities; ~1.65M km²	Project
2022	So2Sat POP	So2Sat POP…	Population benchmark	1.38M multimodal patches; 98 European cities; ~101 GB	Dataset
2022	SEN12MS-CR-TS	SEN12MS-CR-TS…	Multimodal multitemporal cloud-removal dataset	53 global ROIs; >80,000 km²; 30 paired observations/ROI; 40 train + 13 test	Paper
2022	DynamicEarthNet	DynamicEarthNet…	Semantic-change benchmark	75 global AOIs; daily Planet imagery; monthly pixel-wise labels; 7 LULC classes	Paper
2021	DENETHOR	DENETHOR: The DynamicEarthNET Dataset for Harmonized, Inter-Operable, Analysis-Ready, Daily Crop Monitoring from Space	Daily multimodal crop dataset	~4,500 fields; 9 crop classes; daily 3 m Planet Fusion + S1/S2; 2 × 24×24 km train/test areas; ~3 TB	Repository
2021	Million-AID	On Creating Benchmark Dataset for Aerial Image Interpretation…	Large-scale aerial benchmark	~1M aerial images; global coverage; 0.5–11.4 m	Project
2021	S2FL Multimodal Benchmarks	Multimodal Remote Sensing Benchmark Datasets for Land Cover Classification…	3 multimodal benchmark datasets	Houston2013 (HSI+MSI), Berlin (HSI+SAR), Augsburg (HSI+SAR+DSM)	Repository
2021	ERA	ERA: A Dataset and Deep Learning Benchmark for Event Recognition in Aerial Videos	Aerial-video benchmark	2,864 videos; 25 event classes; 5 s clips	Project/Data
2020/21	SEN12MS-CR	Multi-Sensor Data Fusion for Cloud Removal…	Cloud-removal dataset	122,218 S1/cloudy-S2/cloud-free-S2 triplets; 175 global ROIs; 4 seasons	Dataset
2020	So2Sat LCZ42	So2Sat LCZ42…	Global LCZ benchmark	400,673 S1/S2 pairs; 42 cities + additional areas; 17 LCZ classes	Dataset
2019	SEN12MS	SEN12MS – A Curated Dataset…	Global multimodal dataset	180,662 S1/S2/MODIS triplets; 4 seasons; ~421 GiB	Dataset
2018	SEN1-2	The SEN1-2 Dataset for Deep Learning in SAR-Optical Data Fusion	SAR–optical benchmark	282,384 co-registered SAR–optical pairs	Paper
2018	Building Instance / Street-View Benchmark	Building Instance Classification Using Street View Images	Street-view benchmark	19,658 geotagged 512×512 images; 8 classes; 17,600 train + 2,058 test	Repository
2017	SARptical	Fusing Meter-Resolution 4-D InSAR Point Clouds and Optical Images…	SAR–optical multimodal dataset	Public co-registered SAR/InSAR–optical data; exact compact scale metric not clearly stated	Repository
2016	J-SparseFI Data	Exploiting Joint Sparsity for Pan-sharpening – the J-SparseFI Algorithm	Reproducibility/research data	Public experimental data package;	Data Link
2016	IEEE GRSS Data Fusion Contest – Space Video GT	Spatiotemporal Scene Interpretation of Space Videos…	Challenge / ground truth	Public contest ground truth;	Data Link