Earth observation
PulseAugur coverage of Earth observation — every cluster mentioning Earth observation across labs, papers, and developer communities, ranked by signal.
9 day(s) with sentiment data
-
GeoForge framework enhances Earth observation agent reasoning
Researchers have introduced GeoForge, a novel framework designed to enhance the reasoning capabilities of Earth observation (EO) agents. This training-free system self-evolves by converting completed tool-use trajectori…
-
Onboard VLMs enable bandwidth-efficient Earth observation via dialogue
Researchers have developed a novel "Summarize First, Download Later" paradigm for Earth observation satellites, utilizing onboard Vision-Language Models (VLMs) to address bandwidth limitations. This approach involves th…
-
New ELMZip framework uses AI for efficient satellite image compression
Researchers have developed ELMZip, a new framework for onboard satellite image compression utilizing Extreme Learning Machines (ELMs). This method addresses the challenge of transmitting large volumes of data from small…
-
New DARAD framework enhances continual remote sensing image-text retrieval
Researchers have developed DARAD, a novel framework designed to improve continual remote sensing image-text retrieval. This method addresses challenges posed by evolving data archives, such as scale variation and distri…
-
HALO framework enhances low-light remote sensing images by overcoming attention drift · 2 sources tracked
Researchers have developed HALO, a novel framework designed to enhance remote sensing images degraded by extreme low-light conditions. This framework addresses the issue of "attention drift" in existing methods, which l…
-
New chapter details "Earth Embeddings" for satellite imagery analysis
A new chapter on "Earth Embeddings" has been published on arXiv, detailing how earth observation is shifting towards reusable data products rather than requiring users to run large foundation models themselves. These em…
-
EfficientViT-M2 leads in robust onboard satellite image classification
A comparative study evaluated 14 different computer vision models, including various Vision Transformer (ViT) architectures, for onboard satellite image classification in Earth observation tasks. The research focused on…
-
GeoCore-9B: New generative model for Earth observation trained on geospatial data
Researchers have introduced GeoCore-9B, a new 9-billion-parameter generative foundation model specifically designed for Earth observation tasks. Unlike previous models that fine-tuned natural image priors, GeoCore-9B is…
-
SPECTRA framework enhances geospatial model fine-tuning with band routing and efficient LoRA
Researchers have introduced SPECTRA, a novel framework designed to enhance the fine-tuning of geospatial foundation models (GeoFMs) for downstream tasks. SPECTRA addresses two key challenges: spectral mismatch, where do…
-
Quantum CNNs explored for volcanic cloud detection in satellite imagery
Researchers have explored the use of Quantum Convolutional Neural Networks (QCNNs) for detecting volcanic clouds in multispectral satellite imagery. These hybrid models integrate quantum computational layers into classi…
-
Hybrid Quantum CNN enhances volcanic thermal activity recognition
Researchers have developed a novel Hybrid Quantum AlexNet architecture designed to improve the recognition of volcanic thermal activity from satellite imagery. This model integrates a classical convolutional neural netw…
-
New benchmark evaluates multimodal models for real-time disaster intelligence
Researchers have introduced Obshazard-bench, a new benchmark designed to evaluate how well multimodal foundation models can process raw Earth observation data for real-time disaster intelligence. Unlike existing benchma…
-
New OVEarth-Bench benchmark evaluates open-vocabulary Earth observation models
A new benchmark called OVEarth-Bench has been introduced to evaluate open-vocabulary Earth observation capabilities. This benchmark addresses limitations in existing evaluations by expanding category breadth and query d…
-
Paper details best practices for ML-driven geospatial map production
A new paper outlines best practices for creating large-scale geospatial map products using machine learning and Earth observation data. The paper addresses challenges in the end-to-end pipeline, from data preprocessing …
-
Remote sensing VLM "More with Less" prioritizes data scale over architecture
Researchers have developed a large-scale remote sensing vision-language model (VLM) called "More with Less" that challenges the need for specialized architectural designs. By training a general-purpose VLM on a diverse …
-
SAM 3 evaluation reveals limitations in remote sensing segmentation
A new paper evaluates the capabilities of Segment Anything Model 3 (SAM 3) for remote sensing tasks, finding that while it avoids overfitting and performs well in segmentation, it struggles with sub-pixel resolution and…
-
New research explores efficient adaptation of foundation models for Earth observation
Two recent arXiv papers explore the adaptation and application of foundation models for Earth observation (EO). The first paper discusses design principles for remote sensing foundation models (RSFMs), emphasizing domai…
-
TESSERA v2 study reveals optimal scaling for Earth-observation models
Researchers have conducted a large-scale study on scaling pixel-wise Earth-observation foundation models, involving 395 training runs on 1,024 NVIDIA GH200 superchips. The study found that pretraining loss is a poor pre…
-
New framework adapts 3D models for satellite multi-view reconstruction
Researchers have developed EO-VGGT, a new framework designed to adapt existing 3D foundation models for satellite multi-view reconstruction. This framework addresses the geometric discrepancies between implicit perspect…
-
New SAMBA model enhances SAR target recognition with Mamba and scatter-guided masking
Researchers have introduced SAMBA, a novel foundation model designed for Synthetic Aperture Radar (SAR) target recognition. SAMBA utilizes a Mamba encoder to address the computational complexity of traditional Transform…