Introduces ChronoEarth-492K, a 492K-patch temporally calibrated hyperspectral dataset from the EO-1 Hyperion archive spanning 2001-2017, plus a benchmark for static, short-horizon, and long-horizon spatiotemporal tasks using open geospatial products.
Neural plasticity- inspired foundation model for observing the Earth crossing modalities.arXiv:2403.15356,
25 Pith papers cite this work, alongside 21 external citations. Polarity classification is still indexing.
25
Pith papers citing it
21
external citations · external index
citation-role summary
background 3
citation-polarity summary
roles
background 3polarities
background 3representative citing papers
Introduces the CHC dataset of 3 m continuous canopy height differences with uncertainties and the uncertainty-aware change regression task for fine-tuning GFMs on PlanetScope time series.
Unbalanced optimal transport with self-correction on noisy LiDAR supervision enables effective tree counting from satellite imagery on a new benchmark.
UniverSat is a ViT-style model with a universal patch encoder enabling self-supervised training on heterogeneous multimodal Earth observation data from varying resolutions and sensors.
EarthShift is a new benchmark using paired datasets to measure robustness of geospatial foundation models to realistic distribution shifts, finding consistent 15-20% performance drops out-of-distribution across 8 models and 11 tasks.
Tempov is a self-supervised satellite foundation model that predicts wealth levels and decadal changes at high resolution across Africa from Landsat imagery, outperforming baselines even with limited labels and generalizing temporally.
OmniGCD trains a Transformer once on synthetic data to enable zero-shot generalized category discovery across 16 datasets in four modalities without any dataset-specific fine-tuning.
UNIGEOCLIP creates a unified embedding for aerial imagery, street views, elevation, text, and coordinates via all-to-all contrastive alignment plus a scaled lat-long encoder, outperforming single-modality and coordinate baselines on geospatial tasks.
COP-GEN models multimodal Copernicus Earth observation data as conditional distributions via a latent diffusion transformer, producing diverse physically consistent outputs and covering 90% of the real observation manifold on a new stochastic benchmark.
DeluluNet enables continued prediction under modality substitution, addition, or subsets by training a multi-modal model from a unimodal teacher and unlabeled multimodal data via modality hallucination.
T-SAR-JEPA forecasts latent states from SAR amplitude time series with a domain-adapted ViT and temporal transformer to detect anomalies, reaching 77% ROC-AUC on a volcanic dataset while treating InSAR coherence as independent validation.
SpectralEarth-FM is a multisensor hierarchical transformer pretrained on a 40TB co-located HSI-MSI-SAR dataset using a JEPA-style objective and reports state-of-the-art results on hyperspectral and standard EO benchmarks.
LESSViT introduces a low-rank efficient spatial-spectral attention mechanism and a hyperspectral masked autoencoder to improve generalization across spectral configuration shifts in hyperspectral imagery.
An audit of 152 papers reveals that geospatial foundation models lack standardized evaluations, training controls, and weight releases, so no one knows the state of the art.
Introduces the SMART-HC-VQA dataset with 65k single-image and 2.3M temporal VQA examples plus an adapted LLaVA-NeXT MLLM framework for geospatial-temporal sensemaking of remote sensing construction activity.
CAFe-DINO achieves SOTA open-vocabulary semantic segmentation on remote sensing datasets by leveraging DINOv3 features with cost aggregation and upsampling, fine-tuned solely on an RS-targeted COCO-Stuff subset.
CoDe-MAE achieves effective joint pretraining of heterogeneous high-resolution optical and SAR images via optical-anchored knowledge distillation, conditioned contrastive learning, and cross-modal degraded reconstruction, yielding better data efficiency and downstream performance than larger-scale基础
EFDiff conditions a diffusion model with Prithvi-EO-2.0 geospatial embeddings via cross-attention to achieve 32x LST super-resolution, outperforming baselines on a global Landsat dataset.
TESSERA learns robust label-efficient embeddings from irregular multi-modal EO time series via Barlow Twins plus global shuffling and mix-based regularizers, delivering SOTA accuracy on classification, segmentation and regression tasks while releasing planetary-scale embeddings and code.
Standardized pretraining and evaluation of geospatial multimodal foundation models on GEOBench reveals design trade-offs in flexibility, modality alignment, and task performance.
HQ-JEPA combines JEPA-style predictive self-supervision with cross-modal alignment and a SWAP-test-based quantum fidelity loss for learning representations from paired remote sensing imagery, reporting competitive results on GeoBench tasks.
A generative compression model using historical priors for Earth observation data achieves up to 10,000x reduction after exascale training on an Armv9 supercomputer.
LIANet encodes multi-temporal Earth observation data into a coordinate-based neural field that supports label-only fine-tuning for downstream tasks without access to raw imagery.
A systematic review that introduces a framework for feature extraction in remote sensing, traces its evolution in the data value chain, and synthesizes trends toward unified representations and foundation models.
citing papers explorer
-
ChronoEarth-492K: A Large Scale and Long Horizon Spatiotemporal Hyperspectral Earth Observation Dataset and Benchmark
Introduces ChronoEarth-492K, a 492K-patch temporally calibrated hyperspectral dataset from the EO-1 Hyperion archive spanning 2001-2017, plus a benchmark for static, short-horizon, and long-horizon spatiotemporal tasks using open geospatial products.
-
Uncertainty-aware tree height change regression
Introduces the CHC dataset of 3 m continuous canopy height differences with uncertainties and the uncertainty-aware change regression task for fine-tuning GFMs on PlanetScope time series.
-
Counting Trees from Satellite Imagery with Noisy Supervision
Unbalanced optimal transport with self-correction on noisy LiDAR supervision enables effective tree counting from satellite imagery on a new benchmark.
-
UniverSat: Resolution- and Modality-Agnostic Transformers for Earth Observation
UniverSat is a ViT-style model with a universal patch encoder enabling self-supervised training on heterogeneous multimodal Earth observation data from varying resolutions and sensors.
-
EarthShift: a benchmark for measuring robustness to real-world distribution shifts in Earth observation
EarthShift is a new benchmark using paired datasets to measure robustness of geospatial foundation models to realistic distribution shifts, finding consistent 15-20% performance drops out-of-distribution across 8 models and 11 tasks.
-
A satellite foundation model for improved wealth monitoring
Tempov is a self-supervised satellite foundation model that predicts wealth levels and decadal changes at high resolution across Africa from Landsat imagery, outperforming baselines even with limited labels and generalizing temporally.
-
OmniGCD: Abstracting Generalized Category Discovery for Modality Agnosticism
OmniGCD trains a Transformer once on synthetic data to enable zero-shot generalized category discovery across 16 datasets in four modalities without any dataset-specific fine-tuning.
-
UNIGEOCLIP: Unified Geospatial Contrastive Learning
UNIGEOCLIP creates a unified embedding for aerial imagery, street views, elevation, text, and coordinates via all-to-all contrastive alignment plus a scaled lat-long encoder, outperforming single-modality and coordinate baselines on geospatial tasks.
-
COP-GEN: Latent Diffusion Transformer for Copernicus Earth Observation Data
COP-GEN models multimodal Copernicus Earth observation data as conditional distributions via a latent diffusion transformer, producing diverse physically consistent outputs and covering 90% of the real observation manifold on a new stochastic benchmark.
-
Changing Modalities: Adapting Remote Sensing Models to New Satellites and Sensors
DeluluNet enables continued prediction under modality substitution, addition, or subsets by training a multi-modal model from a unimodal teacher and unlabeled multimodal data via modality hallucination.
-
T-SAR-JEPA: Self-Supervised Temporal Anomaly Detection in SAR Amplitude Stacks via Latent Prediction
T-SAR-JEPA forecasts latent states from SAR amplitude time series with a domain-adapted ViT and temporal transformer to detect anomalies, reaching 77% ROC-AUC on a volcanic dataset while treating InSAR coherence as independent validation.
-
SpectralEarth-FM: Bringing Hyperspectral Imagery into Multimodal Earth Observation Pretraining
SpectralEarth-FM is a multisensor hierarchical transformer pretrained on a 40TB co-located HSI-MSI-SAR dataset using a JEPA-style objective and reports state-of-the-art results on hyperspectral and standard EO benchmarks.
-
LESSViT: Robust Hyperspectral Representation Learning under Spectral Configuration Shift
LESSViT introduces a low-rank efficient spatial-spectral attention mechanism and a hyperspectral masked autoencoder to improve generalization across spectral configuration shifts in hyperspectral imagery.
-
No One Knows the State of the Art in Geospatial Foundation Models
An audit of 152 papers reveals that geospatial foundation models lack standardized evaluations, training controls, and weight releases, so no one knows the state of the art.
-
Geospatial-Temporal Sensemaking of Remote Sensing Activity Detections with Multimodal Large Language Model
Introduces the SMART-HC-VQA dataset with 65k single-image and 2.3M temporal VQA examples plus an adapted LLaVA-NeXT MLLM framework for geospatial-temporal sensemaking of remote sensing construction activity.
-
DINO Soars: DINOv3 for Open-Vocabulary Semantic Segmentation of Remote Sensing Imagery
CAFe-DINO achieves SOTA open-vocabulary semantic segmentation on remote sensing datasets by leveraging DINOv3 features with cost aggregation and upsampling, fine-tuned solely on an RS-targeted COCO-Stuff subset.
-
Better with Less: Tackling Heterogeneous Multi-Modal Image Joint Pretraining via Conditioned and Degraded Masked Autoencoder
CoDe-MAE achieves effective joint pretraining of heterogeneous high-resolution optical and SAR images via optical-anchored knowledge distillation, conditioned contrastive learning, and cross-modal degraded reconstruction, yielding better data efficiency and downstream performance than larger-scale基础
-
When Earth Foundation Models Meet Diffusion: An Application to Land Surface Temperature Super-Resolution
EFDiff conditions a diffusion model with Prithvi-EO-2.0 geospatial embeddings via cross-attention to achieve 32x LST super-resolution, outperforming baselines on a global Landsat dataset.
-
TESSERA: Temporal Embeddings of Surface Spectra for Earth Representation and Analysis
TESSERA learns robust label-efficient embeddings from irregular multi-modal EO time series via Barlow Twins plus global shuffling and mix-based regularizers, delivering SOTA accuracy on classification, segmentation and regression tasks while releasing planetary-scale embeddings and code.
-
Emerging Flexible Designs for Geospatial Multimodal Foundation Models
Standardized pretraining and evaluation of geospatial multimodal foundation models on GEOBench reveals design trade-offs in flexibility, modality alignment, and task performance.
-
HQ-JEPA: Hybrid Quantum Joint-Embedding Predictive Architecture for Cross-Modal Remote Sensing Representation Learning
HQ-JEPA combines JEPA-style predictive self-supervision with cross-modal alignment and a SWAP-test-based quantum fidelity loss for learning representations from paired remote sensing imagery, reporting competitive results on GeoBench tasks.
-
Transforming the Use of Earth Observation Data: Exascale Training of a Generative Compression Model with Historical Priors for up to 10,000x Data Reduction
A generative compression model using historical priors for Earth observation data achieves up to 10,000x reduction after exascale training on an Armv9 supercomputer.
-
Location Is All You Need: Continuous Spatiotemporal Neural Representations of Earth Observation Data
LIANet encodes multi-temporal Earth observation data into a coordinate-based neural field that supports label-only fine-tuning for downstream tasks without access to raw imagery.
-
Feature Extraction in the Remote Sensing Data Value Chain: A Systematic Review of Methods and Applications
A systematic review that introduces a framework for feature extraction in remote sensing, traces its evolution in the data value chain, and synthesizes trends toward unified representations and foundation models.
-
Harvesting AlphaEarth: Benchmarking the Geospatial Foundation Model for Agricultural Downstream Tasks
AEF embeddings perform competitively with RS models for local agricultural tasks but show limited spatial transferability, time sensitivity, and interpretability.