datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
EarthView
EarthView dataset
Overview
The EarthView Dataset is a comprehensive collection of multispectral earth imagery. The dataset is divided into four distinct subsets sourced from Satellogic, Sentinel-1, Sentinel-2, and NEON imagers, each providing unique data.
The dataset is also available in AWS Open Data registry.
And you can play and navigate Satellogic's dataset in this Colab notebook.
Dataset Viewer
Check the EarthView Dataset Viewer and it's… See the full description on the dataset page: https://huggingface.co/datasets/satellogic/EarthView.NatureLM-audio-training
Dataset card for NatureLM-audio-training
Overview
NatureLM-audio-training is a large and diverse audio-language dataset designed for training bioacoustic models that can generate a natural language answer to a natural language query on a reference bioacoustic audio recording.
For example, for an in-the-wild audio recording of a bird species, a relevant query might be "What is the common name for the focal species in the audio?" to which an audio-language model trained… See the full description on the dataset page: https://huggingface.co/datasets/EarthSpeciesProject/NatureLM-audio-training.Earth-BenchGAMUSThe Pytorch dataloader for GAMUS can be found here: https://github.com/EarthNets/RSI-MMSegmentation.
earth2studio-assets
Earth2Studio Repository Assets
Collection of assets used to support the Earth2Studio github repository.
license: apache-2.0
earth-tensorsEarthDial-Dataset
🌍 EarthDial-Dataset
The EarthDial-Dataset is a curated collection of evaluation-only datasets focused on remote sensing and Earth observation downstream tasks. It is designed to benchmark vision-language models (VLMs) and multimodal reasoning systems on real-world scenarios involving satellite and aerial imagery.
📚 Key Features
Evaluation-focused: All datasets are for inference/testing only — no train/val splits.
Diverse Tasks:
Classification
Object Detection
Change… See the full description on the dataset page: https://huggingface.co/datasets/akshaydudhane/EarthDial-Dataset.research10-archivehello this is the fucture
usgs-global-earthquake-catalog
USGS Global Earthquake Catalog
Provides historical data on global seismic events, sourced directly from the U.S. Geological Survey (USGS) Earthquake Hazards Program via its FDSN Event Web Service.
Each record represents a single seismic event (primarily earthquakes) and contains detailed information, including:
Event Time & Location: Precise timestamp, geographic coordinates (latitude, longitude), and depth of the event.
Magnitude: The magnitude of the event (mag) and the method… See the full description on the dataset page: https://huggingface.co/datasets/mnemoraorg/usgs-global-earthquake-catalog.XL_PDXL_Embeddings
SDXL & PDXL Embeddings
Most of these are already on Civitai, and are made with embedding merge. We've converted as many as possible to safetensors and uploaded.
About & Links
About Us
We are the Duskfall Portal Crew, a DID system with over 300 alters, navigating life with DID, ADHD, Autism, and CPTSD. We believe in AI’s potential to break down barriers and enhance mental health, despite its challenges. Join us on our creative journey exploring identity and… See the full description on the dataset page: https://huggingface.co/datasets/EarthnDusk/XL_PDXL_Embeddings.BEANS-Zero
BEANS-Zero
Version: 0.1.0
Created on: 2025-04-12
Creators:
Earth Species Project (https://www.earthspecies.org)
Overview
BEANS-Zero is a bioacoustics benchmark designed to evaluate multimodal audio-language models in zero-shot settings. Introduced in the paper NatureLM-audio paper (Robinson et al., 2025), it brings together tasks from both existing datasets and newly curated resources.
The benchmark focuses on models that take a bioacoustic audio input (e.g., bird or… See the full description on the dataset page: https://huggingface.co/datasets/EarthSpeciesProject/BEANS-Zero.Embeddings_SD15
Positive & Negative TI/Embeddings for Stable Diffusion
Embeddings and Textual Inversions we've made LARGELY using Embedding Merge for Automatic1111, be aware IF A FILE causes you a SINGLETON/TENSOR MUST MACH just yeet it and try another.
About & Sponsored
Want to see more? We're starting to release EXCLUSIVE Content via our patreon: https://patreon.com/earthndusk
"WE"? - We have Dissociative identity disorder, ADHD, Autism and CPTSD - "WE" as in we're a system of over 200… See the full description on the dataset page: https://huggingface.co/datasets/EarthnDusk/Embeddings_SD15.EarthworkSat-7K
EarthworkSat-7K
A satellite imagery dataset for finding earthwork on construction sites, and for estimating its volume.
About 7,400 annotated earthwork instances in 14 cities across Europe and Asia. The images are Google Earth screen captures at a pixel size of about 0.13 m, taken between 2018 and 2025. A separate set of 505 instances at 14 further locations is paired with public airborne LiDAR, so each of those instances has a reference volume in cubic metres.
What… See the full description on the dataset page: https://huggingface.co/datasets/issai/EarthworkSat-7K.EarthVLSetEarthVL: A Progressive Earth Vision-Language Understanding and Generation Framework
by Junjue Wang,
Yanfei Zhong,
Zihang Chen,
Zhuo Zheng,
Ailong Ma, and Liangpei Zhang
[Paper],
[Dataset]
News
2026/01/06, New Global-LoveDA !!! We released the global-scale segmentation leaderboard at Global-LoveDA. Just zip all the test images into one file and submit it.
2026/01/06, The segmentation data is released at [Dataset].
2026/01/06, We are preparing the code and data for… See the full description on the dataset page: https://huggingface.co/datasets/Kingdrone-Junjue/EarthVLSet.rl-game-traces-middle-earth-shadow-of-war
中土世界：战争之影
This public dataset repository contains gameplay trace data uploaded from E:\中土世界：战争之影.
Contents
Files: 418
Total local size: 275.23 GB
Generated: 2026-06-04T16:07:45+00:00
File Types
.jsonl: 128
.png: 98
.json: 96
.mkv: 32
.txt: 32
.parquet: 31
.db: 1
Notes
This repository may contain gameplay video, Parquet files, JSON/JSONL metadata, and input event logs.
The license is marked as other; review game footage, audio… See the full description on the dataset page: https://huggingface.co/datasets/yinhuankuang/rl-game-traces-middle-earth-shadow-of-war.Latent-Earth
Latent Earth: An Atlas of Architecture in Flux.2
200,000 images of 40,000 places on Earth, each rendered by a single
image model in a single state of its training, with five internal
representations recorded for every image while it was being generated.
Nothing else enters. Each prompt contains only a place's name; no
photographs, no maps, no climate records correct what the model proposes.
This is therefore not a depiction of the world but a probe of the model: a
survey of what… See the full description on the dataset page: https://huggingface.co/datasets/Punktiert/Latent-Earth.Earth-Silver
(ICLR'26) EarthSE: A Benchmark for Evaluating Earth Scientific Exploration Capability of LLMs
Updates/News 🆕
🚩 News (2026-01-26) EarthSE has been accepted by ICLR 2026 🎉.
Abstract
Advancements in Large Language Models (LLMs) drive interest in scientific applications, necessitating specialized benchmarks such as Earth science. Existing benchmarks either present a general science focus devoid of Earth science specificity or cover isolated… See the full description on the dataset page: https://huggingface.co/datasets/ai-earth/Earth-Silver.EarthVerse
Benchmarking scientific agents across dynamic Earth systems and natural hazards
Zhiqing Cui1, Xinxiang Yin2, Yihong Tang3, Xinglang Zhang4, Yuanzhe Hu5, Siru Zhong4, Weidong Tang6,
Yuxuan Liang4, Weijia Li7, Ming Jin8, Shirui Pan8, Yuhao Kang9, Dingyi Zhuang10,†, Jinhua Zhao10
1NUIST 2HKU 3McGill 4HKUST(GZ) 5Georgia Tech 6NUS 7Tsinghua 8Griffith 9UT Austin 10MIT †Corresponding author
Project page ·… See the full description on the dataset page: https://huggingface.co/datasets/miracle10/EarthVerse.EarthMind-dataopen_earth_mapEarthScience-MLLM-20K
EarthScience-MLLM-20K
A unified JSONL package for multimodal large-model training across three Earth-science domains:
Meteorology from ZhanxiangHua/WeatherQA_SFT.
Geography / map QA from HuggingFaceM4/the_cauldron config mapqa.
Remote-sensing common-sense QA + grounding/detection from xiang709/VRSBench.
The package intentionally excludes segmentation-style targets. Each JSONL line is one training/evaluation unit.
Files
train.jsonl: 20000 examples.
test.jsonl:… See the full description on the dataset page: https://huggingface.co/datasets/moTcream/EarthScience-MLLM-20K.earth2studio-assetsdynamic_earthnetDynamic EarthNet dataset redistributed from https://mediatum.ub.tum.de/1650201 and https://cvg.cit.tum.de/webshare/u/toker/dynnet_training_splits/ under a common tarball for simpler download speeds.
Individual zip files were replaced with tarballs instead.
In the mediatum server version the following directories have the wrong name compared to the given split txt files:
/labels/5111_4560_13_38S
/labels/6204_3495_13_46N
/labels/7026_3201_13_52N
/labels/7367_5050_13_54S
/labels/2459_4406_13_19S… See the full description on the dataset page: https://huggingface.co/datasets/torchgeo/dynamic_earthnet.earth-sst-dailyearthquakes-daily
Earthquakes Daily — M4.5+ global snapshot
A daily snapshot of every magnitude 4.5+ earthquake in the live USGS feed, exported
through Dynamic Feed — a live, verifiable data API whose every
response is Ed25519-signed. One file per day (data/YYYY-MM-DD.jsonl), one JSON
object per earthquake per line.
Live source: https://dynamicfeed.ai (tool: earthquakes, endpoint POST /v1/batch) — keyless, no signup
Upstream source: USGS Earthquake Hazards Program
Licence: US public domain (USGS… See the full description on the dataset page: https://huggingface.co/datasets/dynamicfeed/earthquakes-daily.EarthSynth-180K
EarthSynth-180K Dataset
EarthSynth-180K is a multi-task, conditional, diffusion-based generative dataset designed for remote sensing image synthesis and understanding.It was introduced in the paper "EarthSynth: Generating Informative Earth Observation with Diffusion Models" (arXiv 2025).
This dataset supports text-to-image generation, mask-conditioned synthesis, and multi-category augmentation for Earth observation research.
Dataset Details… See the full description on the dataset page: https://huggingface.co/datasets/jaychempan/EarthSynth-180K.earthquakes
California Earthquakes 1967-2018
Location, magnitude and type of 2.5+ magnitude earthquakes in California from 1967 to 2018.
Source: https://kepler.gl/
EarthMind-Benchlucas-mega
LUCAS-MEGA
LUCAS-MEGA: A Large-Scale Multimodal Dataset for Representation Learning in Soil-Environment Systems
Manuscript
Introduction
LUCAS-MEGA is a large-scale multimodal dataset for soil-environment systems, built by fusing heterogeneous European soil
and environmental datasets with the LUCAS soil survey as the backbone.
The released dataset contains:
72,000+ soil samples
1,000+ fused soil and environmental features
68 integrated ESDAC source datasets… See the full description on the dataset page: https://huggingface.co/datasets/earthroverprogram/lucas-mega.NASA_Nearest_Earth_Objects_1910-2024CONTEXT:
There are many dangerous bodies in space, one of them is N.E.O. - "Nearest Earth Objects". Some such bodies really pose a danger to the planet Earth, NASA classifies them as "is_hazardous". This dataset contains ALL NASA observations of similar objects from 1910 to 2024!!!
There are 338,199 records of N.E.O. in the Dataset!
Try to predict "is_hazardous" as accurately as possible! (otherwise we will not be ready for an asteroid attack)
SOURCES:
NASA Open API: https://api.nasa.gov/… See the full description on the dataset page: https://huggingface.co/datasets/IvanSher/NASA_Nearest_Earth_Objects_1910-2024.