Machine Learning for Geospatial 🌍

Welcome to the Machine Learning for Geospatial repository! This is a curated collection of resources, tools, programming libraries, and courses dedicated to applying machine learning techniques to geospatial data. Whether you're a beginner or an experienced practitioner, this repository aims to provide everything you need to explore the fascinating intersection of machine learning and geospatial science.

Why Geospatial Machine Learning?

Geospatial data is becoming increasingly important in solving real-world problems like urban planning, disaster management, environmental monitoring, and logistics optimization. By integrating machine learning, we can:

Automate land cover classification from satellite imagery 🌳.
Predict climate trends and weather anomalies 🌦️.
Optimize routes for logistics and transportation 🚚.
Monitor and analyze urban growth 🏙️.

📚 Resources

Beginner Guides

Geospatial Data Basics
- What is GIS?
- Understanding Raster and Vector Data
Intro to Machine Learning
- Machine Learning Crash Course by Google
- Kaggle’s Intro to Machine Learning

Geospatial Data Repositories

Bharatmaps: Bharatmaps.gov.in
ISRO-Bhuvan: Bhuvan-IndiaGeo-Platform of ISRO
OpenStreetMap: OpenStreetMap.org
USGS Earth Explorer: earthexplorer.usgs.gov
Sentinel Hub: Copernicus Sentinel Data
NASA Earth Science Data: earthdata.nasa.gov

🗂️ HuggingFace Datasets Hub

Hugging Face now lists 272+ geospatial-ready datasets. The tables below highlight 15 heavily requested corpora that pair well with the models in this repo--each one publishes clear licenses (Apache-2.0 or MIT) and ready-to-stream splits.

Multi-Modal Earth Observation Collections

Dataset	Modalities & Volume	Why it matters	Link
TerraMesh 10m Cubes	Sentinel-2 RGB + Sentinel-1 VH/VV + Copernicus DEM, global 10 m tiles	Harmonized cubes for training multi-sensor foundation models out-of-the-box.	TerraMesh
FLAIR-HUB	63B annotated pixels across 19 land-cover classes	High-resolution labels engineered for segmentation and active-learning workflows.	FLAIR-HUB
Sen1Floods11	SAR + optical time pairs with flood masks	De facto benchmark for rapid flood mapping and transfer to disaster regions.	Sen1Floods11

Global Classification Benchmarks

Dataset	Samples	Classes & Notes	Link
EuroSAT	27K multispectral chips (Sentinel-2)	10 land-use classes; useful for lightweight transfer checks.	EuroSAT
BigEarthNet	590K image patches (S1 + S2)	43 multi-label land-cover tags, curated for multi-spectral classification.	BigEarthNet
RESISC45	31.5K RGB aerial tiles	45 scene categories with strong baselines for ViT/CNN adapters.	RESISC45
AID	10K aerial chips	30 scene classes, balanced for quick benchmarking.	AID
UC Merced	2.1K high-res RGB tiles	21 classes; perfect sanity check before scaling up.	UC Merced

Segmentation & Detection Corpora

Dataset	Focus	Highlights	Link
SpaceNet-8	Building footprints + roads in multi-sensor stacks	Includes SAR+EO pairs, coastline ports, and routing-ready metadata.	SpaceNet-8
xView	1M labeled objects, 60 categories	One of the largest object-detection datasets for HR imagery.	xView
LEVIR-CD	Bitemporal change detection	637 paired tiles with pixel-level change masks (buildings & infrastructure).	LEVIR-CD

Climate & Environmental Records

Dataset	Era	Highlights	Link
ERA5-Land	1940–present reanalysis	Hourly atmospheric + land diagnostics, ready for climate downscaling.	ERA5
MERRA-2	1980–present, 160+ variables	NASA reanalysis with aerosol and chemistry products for ESG analytics.	MERRA-2

Specialized Collections

Dataset	Specialty	Why you'd use it	Link
ForestNet	Tropical deforestation monitoring	Combines Landsat/Sentinel stacks with expert forest-change labels.	ForestNet
CropHarvest	Global cropland classification	90+ crops with extensive metadata for transfer to regional ag programs.	CropHarvest
SSL4EO-S12	Self-supervised Sentinel-1/2	Pretext dataset for contrastive/MAE training on unlabeled EO swaths.	SSL4EO-S12

from datasets import load_dataset

# Load paired SAR + optical tensors plus labels in a single call
ds = load_dataset(
    "ibm-nasa-geospatial/sen1floods11",
    split="train",
    streaming=True,  # hug the Hub without needing full downloads
)

sample = next(iter(ds))
sar = sample["image_sar"]       # numpy array: [2, H, W]
optical = sample["image_optical"]  # numpy array: [12, H, W]
mask = sample["label"]          # 0/1 flood raster
print("Example keys:", sample.keys())

💡 Explore even more curated lists via geospatial dataset search and the community geospatial dataset collection.

Tutorials and Articles

GeoPandas Documentation: geopandas.org
Satellite Image Analysis with Python: Planet Univesity
Intro to Google Earth Engine Python API: Community Tutorial — step-by-step guide to authenticating, loading imagery, and running analyses with the Earth Engine Python client.

import ee

def initialize_ee():
    """Authenticate (if needed) and initialize the Earth Engine Python API."""
    try:
        ee.Initialize()
    except Exception:
        ee.Authenticate()
        ee.Initialize()


initialize_ee()

# Build a cloud-filtered Sentinel-2 collection over San Francisco for June 2023.
collection = (
    ee.ImageCollection("COPERNICUS/S2_SR")
    .filterDate("2023-06-01", "2023-06-30")
    .filterBounds(ee.Geometry.Point(-122.4194, 37.7749))
    .filter(ee.Filter.lt("CLOUDY_PIXEL_PERCENTAGE", 10))
)

# Reduce to a representative median image and compute NDVI.
image = collection.median()
ndvi = image.normalizedDifference(["B8", "B4"]).rename("NDVI")

# Define an export footprint (a simple rectangle around San Francisco).
region = ee.Geometry.Rectangle([-123, 37, -122, 38])

# Kick off a Drive export; monitor progress in the Earth Engine Code Editor.
export_task = ee.batch.Export.image.toDrive(
    image=ndvi,
    description="sf_ndvi_june2023",
    folder="earth-engine",
    fileNamePrefix="sf_ndvi_june2023",
    region=region.getInfo()["coordinates"],
    scale=10,
    maxPixels=1e10,
)
export_task.start()
print("Export started. Check the Earth Engine Tasks tab.")

Hands-On: Basic Concepts

The Notebooks/geopython-tutorials directory now hosts a curated set of hands-on notebooks. Launch any concept directly in Colab using the badge links below.

Topic	Concept	Link to Colab
GeoPandas	Bulk Address Geocoding Workflow
	Spatial Query Playbook
	Targeted Shapefile Extraction
	Fuzzy Table Joins Toolkit
XArray for Raster Data Processing	Raster Styling & Quicklook Analytics
	Seamless Mosaics & AOI Clipping
	Raster Sampling Recipes
	Zonal Stats Automation
	Single-Pixel Time-Series Miner
	Multi-Site Time-Series Matrix
	Satellite Stack Processing Pipeline
XArray for Climate Data Processing	Rainfall Aggregation Dashboard
	Climate Anomaly Finder
	Pixel Trend Explorer
XArray Utilities	Array-to-Raster Forge
XArray Utilities	Longitude Normalizer
Dask for Parallel Processing	Parallel Spatial Data Forge
Dask for Parallel Processing	Dask Median Composite Lab
XEE (XArray + Google Earth Engine)	ImageCollection → NetCDF Exporter
	XEE Time-Series Studio
	XEE Image Downloader
	SPI Calculator with XEE
Segment Geospatial	Segmenting Farm Boundaries
Segment Geospatial	Segmenting Mine Perimeters
Web APIs	ORS Facility Proximity Matrix
	OpenAI News Geocoder
	OpenMeteo Forecast Bot

🛠️ Tools and Programming Libraries

Geospatial Libraries

Library	Purpose	Link
GeoPandas	Handle and analyze vector geospatial data.	GitHub
Rasterio	Read and write raster datasets.	GitHub
Shapely	Perform geometric operations like intersections and unions.	GitHub
Fiona	Read and write vector geospatial data.	GitHub
xarray	Work with multidimensional geospatial datasets.	GitHub

Machine Learning Libraries

Library	Purpose	Link
TorchGeo	Deep learning toolkit tailored for geospatial data.	GitHub
scikit-learn	Classic ML algorithms for classification, regression, and clustering.	GitHub
TensorFlow / PyTorch	Deep learning frameworks for tasks like image segmentation.	TensorFlow · PyTorch
LightGBM / XGBoost	Gradient boosting libraries for tabular geospatial datasets.	LightGBM · XGBoost

Visualization Tools

Tool	Purpose	Link
Leaflet	JavaScript library for mobile-friendly interactive maps.	GitHub
Folium	Build interactive maps from Python.	GitHub
Kepler.gl	Web-based tool for large-scale geospatial visualizations.	GitHub
Deck.gl	High-performance 3D geospatial visualizations.	GitHub

🤖 Geospatial AI Model Hub

Category	Model	Why it matters	References
Foundation & SSL	Prithvi-EO (IBM–NASA)	HLS-pretrained foundation models for broad EO transfer.	Hugging Face · IBM Research · NASA Earthdata
Foundation & SSL	SatMAE	Masked autoencoder tuned for temporal/multispectral imagery.	arXiv · NeurIPS 2022
Foundation & SSL	SeCo (Seasonal Contrast)	Seasonal Sentinel-2 contrastive pretraining; strong downstream lifts.	CVPR 2021
Foundation & SSL	Scale-MAE	Scale-aware MAE that improves SpaceNet-style building segmentation.	arXiv · CVPR 2023
Foundation & SSL	SatlasPretrain (AI2)	Ready-to-use encoders for Sentinel/Landsat and aerial imagery.	GitHub
Foundation & SSL	TorchGeo pretrained suite	Catalog of EO-specific backbones (EuroSAT, So2Sat, etc.).	Docs
Land-use / LULC	EuroSAT pretrained models	CNN/ViT baselines for Sentinel-2 land-use classification.	GitHub
Land-use / LULC	BigEarthNet encoders	Multilabel S1/S2 classifiers for regional land-cover tasks.	BigEarthNet
Building footprints	TernausNet / TernausNetV2	U-Net variants widely used on SpaceNet building segmentation.	arXiv
Building footprints	Mask R-CNN / U-Net baselines	Strong SpaceNet MVOI baselines for instance & semantic footprints.	Medium
Road extraction	CRESI / CRESIv2	End-to-end road network & speed extraction (SpaceNet-5 baseline).	GitHub · CVPRW 2020
Road extraction	SpaceNet-5 baselines	Routing-quality metrics like APLS_time for road graph scoring.	Medium
Change detection	ChangeFormer	Transformer Siamese change detection with open weights.	arXiv · GitHub
Change detection	TUNetCD	Transformer-U-Net hybrid for remote-sensing change maps.	PMC
Cloud masking	s2cloudless	Lightweight Sentinel-2 cloud probability & mask generator.	GitHub
Cloud masking	CloudSEN12 / CloudSEN12+	Benchmark dataset & models for clouds/shadows.	Nature Sci. Data · Project
Cloud masking	CloudS2Mask	DL library for high-accuracy Sentinel-2 cloud/shadow detection.	ScienceDirect
Flood mapping	Sen1Floods11 U-Nets	Benchmarks for SAR/optical flood segmentation.	GitHub
Object detection	xView baselines & YOLO	Large-scale overhead object detection benchmarks.	xView · Ultralytics Docs
SAM for EO	SAMRS	NeurIPS’23 dataset/code for Segment Anything in remote sensing.	GitHub · arXiv
Disaster response	Turkey Building Damage Assessment	Rapid post-quake damage grading using high-res imagery.	Project · Image
Disaster response	Building damage assessment (Siamese CNN)	Siamese CNN for global disaster impact estimation.	Publication · Code · Image
Agricultural infrastructure	Poultry barn mapping	Detects industrial poultry barns to monitor environmental impact.	Publication · Code · Image
Environmental monitoring	Glacier mapping & glacial lakes	Tracks glacier change in the Hindu Kush Himalaya region.	Project · Code · Image
Land-use mapping	Land cover mapping (Microsoft Research)	Country-scale LULC maps with label-scarce deep learning.	Project · Downloads · Image
Renewable energy siting	Renewable Energy Mapping	Identifies solar infrastructure footprints across India.	Publication · Code · Image

🤗 HuggingFace Models Hub

15+ openly licensed (Apache-2.0 or MIT) geospatial models on Hugging Face pair with this repo's workflows. Start with the foundation table, then grab fine-tuned checkpoints or enterprise Granite variants as needed.

Foundation & Base Models

Model	Params	Focus	License	Link
Prithvi-EO-1.0-100M	100M	HLS pretraining across 100+ countries; ideal lightweight encoder for Edge/Colab.	Apache-2.0	HF
Prithvi-EO-2.0-300M	300M	Flagship EO foundation updated with improved atmospheric normalization.	Apache-2.0	HF
Prithvi-EO-2.0-600M	600M	Push-button upgrade for higher fidelity segmentation/classification heads.	Apache-2.0	HF
Prithvi-WxC-1.0-2300M	2.3B	Weather+climate aware backbone using ERA5 + satellite reanalysis streams.	Apache-2.0	HF
Prithvi-EO-2.0-300M-TL	300M	Task library variant with adapters for detection/segmentation notebooks.	Apache-2.0	HF
SatCLIP ViT16-L40	304M	Microsoft global contrastive model for scene-to-text grounding.	MIT	HF
SatCLIP ResNet18	11M	Lightweight SatCLIP encoder for embedded inference.	MIT	HF
Satlas-Pretrain ViT-B	86M	AllenAI multi-resolution pretraining on Sentinel/Landsat.	Apache-2.0	HF

Fine-Tuned Specializations

Model	Task	Notes	Link
Prithvi-EO-2.0-300M-Sen1Floods11	Flood segmentation	SAR+optical encoder-decoder with pixel flood masks.	HF
Prithvi-EO-2.0-300M-Burn-Scar	Wildfire burn scar mapping	Multi-temporal ingestion to capture pre/post-fire signatures.	HF
Prithvi-EO-2.0-300M-CropHarvest	Crop classification	Trained on CropHarvest for agri analytics dashboards.	HF
Prithvi-EO-2.0-300M-Multitemporal-Crops	Time-series crop typing	Adds temporal attention heads for season aware decisions.	HF

IBM Granite for Enterprises

Model	Primary KPI	Notes	Link
Granite-Geo-Biomass	Biomass estimation	Optimized for ESG reporting with uncertainty heads.	HF
Granite-Geo-Land-Cover	Land cover classification	Enterprise-ready model with support contracts and batch APIs.	HF

Vision-Language for Remote Sensing QA

Model	Params	Why it helps	Link
Google PaliGemma-3B-GEO	3B	Combine textual prompts with EO imagery for question answering, captioning, or retrieval.	HF

Quick Access Collections

🛰️ Domain Imagery Gallery

Domain	Snapshot	Source
Agriculture		Microsoft Research – Poultry barn mapping
Weather Forecasting		Generated with Matplotlib (pressure anomaly contours)
Rainfall		Generated with Matplotlib (monthly rainfall estimate)
Land Inspection		Generated with Matplotlib (land inspection grid)

🌍GeoSpatial Tech Stack

Transform

Tool	What it’s for	Icon
GDAL	Core geospatial I/O, reprojection, raster/vector ops	🧰
seer.ai	Data prep/ETL with geo context	🔮
dbt	SQL-based transformations/versioned models	🧱
Airbyte	Connectors for syncing data sources	🔌
BigGeo	Large-scale geo ETL pipelines	🌏📦
A5	Workflow utilities for data engineering	🛠️

Analytics

Tool	What it’s for	Icon
CARTO	Cloud location analytics & APIs	🗺️
Fused	Unified geospatial analytics workspace	🧪
Foursquare (FSQ)	POI, movement, visitation analytics	📍
Hex	Notebooks + apps for data teams	🧮
kepler.gl	High-performance geo viz in browser	🧭
Preset	Managed Apache Superset	🎛️
Apache Superset	BI dashboards, SQL exploration	📊

GIS

Tool	What it’s for	Icon
QGIS	Desktop GIS (edit, analyze, visualize)	🧭🖥️
Felt	Collaborative web GIS mapping	🧩
Atlas	Lightweight map design/hosting	🗺️✨
NextGIS	GIS stack & services	🧱🗺️

AI

Tool	What it’s for	Icon
Klarety	GeoAI insights from messy geo data	🧠🗺️
Monarcha	ML for EO/RS workflows	🦋
CONTOUR	Foundation models for maps/imagery	🗺️🤖
Bunting Labs	Geo dev tools / APIs	🐦
aino	AI copilots for spatial ops	🤖
Mundi	Earth data access + analytics	🌍
GeoRetina	Vision models for remote sensing	👁️🛰️
AskEarth	Natural-language Q&A over Earth data	💬🌎

Python

Tool	What it’s for	Icon
GeoPandas	Vector dataframes & spatial ops	🐼🗺️
rasterio	Raster I/O (GDAL bindings)	🧾🛰️
TorchGeo	PyTorch datasets/models for EO	🔥🌍
leafmap	Jupyter interactive maps	🍃
geemap	Earth Engine + geospatial viz	🗺️🔧
lonboard	WebGL vector map rendering	🧭⚡

Formats

Format	What it’s for	Icon
GeoParquet	Columnar vectors with geo metadata	📦🗺️
COG (Cloud-Optimized GeoTIFF)	HTTP-friendly rasters	☁️🖼️
STAC	Cataloging EO assets	🗂️
COPC.io (point clouds)	Streamable point clouds	☁️📍
Zarr	Chunked n-D arrays (cloud-native)	🧱📚

Apps

Tool	What it’s for	Icon
Mapbox	Maps, tiles, SDKs	🗺️📦
deck.gl	GPU-powered web viz	🧊
MapLibre	Open map renderers (GL/Native)	🗺️🆓
geobase	Location SDK/utilities	🧩📍
Veda	NASA/USG EO visualization	🛰️📽️

Orchestrate

Tool	What it’s for	Icon
Apache Airflow	DAG-based scheduling	🌬️
Astronomer	Managed Airflow	🌠
dagster	Data orchestration & assets	🕸️
kestra	Declarative workflows	🏗️
Prefect	Pythonic orchestration	✅

Processing

Tool	What it’s for	Icon
Wherobots	Cloud-native spatial analytics	🤖🗺️
Coiled	Managed Dask at scale	🌀
Databricks	Lakehouse + Spark ML	🔥💎
Apache Sedona	Spatial SQL on Spark	🗺️🔥
Dask	Parallel Python	🧩⚡
Apache Spark	Distributed compute	🔥

OLAP (Analytical)

Tool	What it’s for	Icon
Snowflake	Cloud data warehouse	❄️
MotherDuck	Serverless DuckDB	🦆👩‍🍼
DuckDB	In-process analytics	🦆
Google BigQuery	Serverless DW + GIS	🔎📦
Amazon Redshift	AWS DW	🟥
Trino	SQL query engine (federation)	🚀

Catalogs / Tables

Tool	What it’s for	Icon
Apache Iceberg	Open table format	🧊
Delta Lake	ACID tables on data lakes	🔺
Earthmover	Geo data cataloging/ETL	🚚🌍
DuckLake	DuckDB-native lakehouse tables	🦆🌊

Cloud Storage

Tool	What it’s for	Icon
Amazon S3	Object storage	🪣
Google Cloud Storage	Object storage	☁️🪣
Azure Storage	Blobs/files/queues	🔷🪣
Wasabi	S3-compatible storage	🌶️🪣
Cloudflare R2 / Backblaze-style	Low-egress object storage	🌀🪣
obstore	S3-compatible OSS	📦🪣

🧭 Geospatial Fundamentals Cheat Sheet

Concept	Why it matters	Learn more
Raster data	Pixel/grid representation of continuous surfaces (imagery, elevation).	ArcGIS Pro — Raster data
Vector data	Points/lines/polygons for discrete features (roads, parcels).	Esri GIS Dictionary — Vector
Raster vs Vector	Know when to use each model for analysis and storage.	GISGeography — Data types
Coordinate Reference System (CRS)	Defines how coordinates map to Earth; essential for accuracy.	PROJ — About CRS
EPSG codes	Standard identifiers for CRSs (e.g., 4326, 3857).	pyproj — CRS reference
WGS84 / Lat-Lon (EPSG:4326)	Global geographic CRS used widely in data exchange.	Wikipedia — EPSG:4326
Web Mercator (EPSG:3857)	Web mapping projection (Google/OSM tiles).	Wikipedia — Web Mercator
UTM	Projected CRS in 6° zones for low-distortion mapping.	Wikipedia — UTM
Map projections (Mercator)	Transform globe to flat map; understand distortion.	Britannica — Mercator projection
GeoTIFF	Georeferenced raster format standard.	OGC — GeoTIFF
Cloud-Optimized GeoTIFF (COG)	GeoTIFF layout optimized for HTTP range reads.	COG — Specification
GeoPackage (.gpkg)	SQLite-based container for vector/rasters/tiles.	OGC — GeoPackage
Shapefile	Legacy vector format (SHP/SHX/DBF triplet).	Esri — Shapefile whitepaper
GeoJSON	JSON-based vector data format/spec.	IETF RFC 7946
MBTiles	SQLite container for tiled maps (raster/vector).	Mapbox — MBTiles spec
GeoParquet	Columnar (Parquet) storage with geospatial metadata.	GeoParquet v1.1.0
NetCDF	Self-describing array format used for climate/EO.	Unidata — NetCDF
HDF5	High-performance hierarchical scientific data format.	The HDF Group — HDF5 intro
GDAL	Core library/CLI for raster & vector I/O and transforms.	GDAL documentation
gdalwarp (reproject/warp)	Reprojection, resampling, mosaicking.	GDAL — gdalwarp
Resampling methods	Nearest, bilinear, cubic, lanczos, average, etc.	GDAL — raster resize
PROJ / pyproj	Geodetic/projection transforms in code.	pyproj — PROJ API
GeoPandas	Pandas + geometry for vector data in Python.	GeoPandas — Introduction
Spatial joins (GeoPandas)	Combine layers by spatial relationships.	GeoPandas — Spatial joins
Shapely	Geometry objects & operations (buffer, dissolve…).	Shapely repository
Rasterio	Python raster I/O built on GDAL.	Rasterio — Docs PDF
Xarray	Labeled N-D arrays (great for EO/gridded time series).	xarray documentation
rioxarray	Geo-enhancements for Xarray (CRS, transform, IO).	rioxarray documentation
PDAL & LAS	Point-cloud processing & LAS point format spec.	PDAL docs · ASPRS — LAS 1.4
WMS (OGC)	Request map images (rendered rasters) via web.	OGC — WMS
WMTS (OGC)	Tiled map service for fast web maps.	OGC — WMTS
WFS (OGC)	Web access to actual vector features.	OGC — WFS
STAC	Common spec to catalog/discover spatiotemporal assets.	STAC specification
Sentinel-2 (MSI)	13-band optical imagery (10/20/60 m).	Copernicus — Sentinel-2
Sentinel-1 (SAR)	C-band radar; day/night, cloud-penetrating.	ESA — Sentinel-1 SAR basics
Landsat program	Long-running US optical EO archive.	USGS — Landsat missions
SRTM DEM	Global elevation (void-filled) dataset.	USGS — SRTM
Copernicus DEM	Global DSM (GLO-30/90), GeoTIFF/DTED.	Copernicus — DEM
NDVI	Vegetation “greenness” index (NIR-Red)/(NIR+Red).	NASA Earthdata — NDVI
NDWI (Gao 1996)	Vegetation water content index (NIR-SWIR)/(NIR+SWIR).	ScienceDirect — NDWI
NBR	Burn severity index (NIR-SWIR)/(NIR+SWIR).	USGS — NBR
Cloud masking (QA60)	Sentinel-2 bitmask for clouds/cirrus (used in GEE).	Google Earth Engine — Sentinel-2 SR
Zonal statistics	Summarize raster values over polygons.	rasterstats — Zonal stats
Spatial autocorrelation (Moran’s I)	Detect clustering/dispersion in spatial data.	GeoDa — Moran’s I
Tobler’s First Law	“Near things are more related than distant things.”	Wikipedia — Tobler’s law
MAUP	Bias from aggregating data into arbitrary zones.	Wikipedia — MAUP
Reprojection	Change dataset CRS to align analyses/maps.	GDAL — raster reproject
Rescaling & resampling	Change raster resolution methodically.	GDAL — raster resize
PostGIS	Spatial SQL in PostgreSQL (geometry/geography + ops).	PostGIS documentation
QGIS	Open-source desktop GIS (editing, analysis, viz).	QGIS user manual
STAC in practice	How agencies expose catalogs (example implementation).	USGS — STAC example

🚀 HuggingFace Spaces & Interactive Apps

Hugging Face Spaces make it easy to demo, duplicate, and productionize geospatial AI without standing up servers. Below are 8+ interactive experiences plus deployment tips that map directly to the models/datasets highlighted above.

IBM-NASA Prithvi Demo Suite

Space	What you can test	Link
Prithvi EO 2.0 Cloud Gap	Fill cloud gaps, compare reconstructions, and download clean tiles.	Space
Sen1Floods11 Flood Segmentation	Upload SAR + optical chips and return pixel-accurate flood masks.	Space
Burn Scar Monitor	Detect wildfire burn scars using bi-temporal Sentinel stacks.	Space
Crop Classification Lab	Multi-temporal crop typing with confidence scores and feature export.	Space
Prithvi EO Explorer	Inspect embeddings, run zero-shot queries, and download features.	Space

Visualization & Analysis Tools

Space	Capability	Link
Granite Geospatial Explorer	Compare Granite biomass/land-cover predictions with ground truth layers.	Space
Global Flood Dashboard	Monitor near real-time flood risk with Prithvi + ERA5 fusion layers.	Space
Remote Sensing Playground	Leafmap/MapLibre-based viewer for overlaying datasets, embeddings, or masks.	Space

📦 Voila collection: ready-to-deploy Jupyter apps for Spaces live in voila-dashboards/voila-gallery. Duplicate a template, swap the notebook, and push directly to any Hugging Face Space.

How to Use or Duplicate a Space

Launch & explore - open the space, switch hardware (CPU/GPU/T4) as needed, and watch the console for preprocessing steps.
Duplicate - click Duplicate Space to clone into your namespace with secrets preserved via the Variables panel.
Access via API - use the Use via API tab for ready-made curl, Python, and JS snippets plus hf_token guidance.
Embed & automate - copy the iframe snippet for docs/notebooks or call the Space endpoint from workflows via huggingface_hub.InferenceClient.

# Create and publish your own geospatial Space
pip install -U "huggingface_hub[cli]"
huggingface-cli login
huggingface-cli repo create space your-hf-handle/prithvi-demo --type gradio --sdk gradio
git clone https://huggingface.co/spaces/your-hf-handle/prithvi-demo
cd prithvi-demo
python - <<'EOF'
from pathlib import Path
Path("app.py").write_text(
    "import gradio as gr\n"
    "def run(image_path):\n"
    "    return image_path\n"
    "iface = gr.Interface(fn=run, inputs=gr.Image(type='filepath'), outputs='image')\n"
    "iface.launch()\n"
)
EOF
huggingface-cli upload --repo-type space --path . your-hf-handle/prithvi-demo
huggingface-cli space hardware set your-hf-handle/prithvi-demo cpu-upgrade

Frameworks & Deployment Options

Framework	Ideal for	Docs
Gradio	Fast MVPs and model cards with sliders/maps.	Docs
Streamlit	Data storytelling dashboards with charts + maps.	Docs
Voila	Turn notebooks into reproducible apps without rewriting code.	Docs
Static HTML + MapLibre	Lightweight viewers that only need tiles/geojson.	Docs

Explore More Spaces

🎓 Courses to Learn Geospatial and Machine Learning

Free Courses

Introduction to GIS
- Esri’s GIS Basics
- QGIS Tutorials
Satellite Image Processing
Machine Learning Essentials
- FastAI’s Deep Learning for Coders (includes image classification tasks that can be applied to geospatial data).
- Machine Learning with Python by Coursera

Paid Courses

Advanced GIS and Remote Sensing
- Coursera: GIS, Mapping, and Spatial Analysis Specialization
- Udemy: Python for Geospatial Analysis
Deep Learning for Geospatial Applications
- Deep Learning for Satellite Imagery by AWS
- Omdena’s Geospatial AI Courses

Machine Learning for Remote Sensing (ML4RS) Workshop @ ICLR 2024 Microsoft Sponsor

This is a curated list of papers presented at the Machine Learning for Remote Sensing (ML4RS) Workshop at ICLR 2023 & 2024.

Title	Presentation Type	Award
Good at Captioning, Bad at Counting: Benchmarking GPT-4V on Earth Observation Data	Oral	Best Paper
A Concise Tiling Strategy for Preserving Spatial Context in Earth Observation Imagery	Poster	Best Poster
Causal Graph Neural Networks for Wildfire Danger Prediction	Oral
Sparsely Labeled Land Cover Classification with Oversegmentation-based Graph U-Nets	Oral
Uncertainty Aware Tropical Cyclone Wind Speed Estimation from Satellite Data	Oral
CromSS: Cross-Modal Pre-Training with Noisy Labels for Remote Sensing Image Segmentation	Oral
Leveraging Synthetic Data and Machine Learning for Cloud Optical Thickness Estimation	Oral
Global Aboveground Biomass Density Estimation from Sentinel-2 Imagery	Poster
AI-Powered School Mapping and Connectivity Status Prediction Using Satellite Imagery	Poster
Evaluating Tool-Augmented Agents in Remote Sensing Platforms	Poster
A Change Detection Reality Check	Poster
Uncertainty Quantification for Probabilistic Machine Learning in Earth Observation	Poster
Red Teaming Models for Hyperspectral Image Analysis Using Explainable AI	Poster

Machine Learning for Remote Sensing (ML4RS) Workshop @ ICLR 2023

Title	Presentation Type
Towards Explainable Land Cover Mapping: A Counterfactual-Based Strategy	Oral
Enhancing Acoustic Classification Using Meta-Data	Oral
Explaining Multimodal Data Fusion: Occlusion Analysis for Wilderness Mapping	Oral
Mask Conditional Synthetic Satellite Imagery	Oral
Optimizing Ecosystem Monitoring through Machine Learning	Oral
Aerial View Localization with Reinforcement Learning: Towards Emulating Search-and-Rescue	Poster
Improved Marine Debris Detection in Satellite Imagery with an Automatic Refinement of Coarse Hand Annotations	Poster
Titan Cloud Identification with Deep Transfer Learning	Poster
Evaluation Challenges for Geospatial ML	Poster
Building Light Models with Competitive Performance for Remote Sensing	Poster
Unsupervised Domain Adaptation for Semantic Segmentation of Dwellings with Unbalanced Optimal Transport	Poster
Efficient Ship Detection on Large Open Sea Areas	Poster
Remote Control: Debiasing Remote Sensing Predictions for Causal Inference	Poster
Improve State-Level Wheat Yield Forecasts in Kazakhstan on GEOGLAM’s EO Data by Leveraging a Simple Spatial-Aware Technique	Poster

📁 Repository Contents [New additions are WIP]

📂 Datasets: Curated datasets for geospatial ML experiments.
📂 Tutorials: Step-by-step guides and example notebooks.
📂 Notebooks: Jupyter notebooks demonstrating geospatial machine learning workflows (1 in-repo quickstart).
📂 Notebooks/earthengine: Vendored Google Earth Engine workflows (cloud masking, classification, change detection) sourced from the community — 42 notebooks across:
- CloudMasking (4) · RasterProcessing (8) · ArrayAnalytics (4) · VectorAndZonal (4) · SpatialJoins (1)
- ImageCollections (3) · Segmentation (1) · Detection (1) · MachineLearning (4)
- Terrain (3) · WaterMonitoring (4) · ChangeMonitoring (2) · Visualization (3)
📂 Tools: Scripts and utilities for geospatial data processing.

🤝 Contributing

We welcome contributions! Whether you want to add a new resource, share a dataset, or submit a tutorial, here’s how you can help:

Fork the repository.
Create a new branch for your changes.
Submit a pull request with a detailed description.

Show your support

Give a 🌟 if this repo helped you!

📞 Contact

For questions, suggestions, or collaborations, feel free to open an issue or connect with me at:

Email: [sache.meet@yahoo.com]
LinkedIn: [www.linkedin.com/in/curious-susant]

@misc{Curious Susant,
    author       = {S Susant Achary},
    title        = {Machine-Learning-for-Geospatial},
    year         = {2025}
}

Name		Name	Last commit message	Last commit date
Latest commit History 37 Commits
Datasets		Datasets
Navigation		Navigation
Notebooks		Notebooks
Tools		Tools
Tutorials		Tutorials
assets		assets
usecases		usecases
.gitattributes		.gitattributes
LICENSE		LICENSE
README.md		README.md

Folders and files

Latest commit

History

Repository files navigation

Machine Learning for Geospatial 🌍

Why Geospatial Machine Learning?

📚 Resources

Beginner Guides

Geospatial Data Repositories

🗂️ HuggingFace Datasets Hub

Multi-Modal Earth Observation Collections

Global Classification Benchmarks

Segmentation & Detection Corpora

Climate & Environmental Records

Specialized Collections

Tutorials and Articles

Hands-On: Basic Concepts

🛠️ Tools and Programming Libraries

Geospatial Libraries

Machine Learning Libraries

Visualization Tools

🤖 Geospatial AI Model Hub

🤗 HuggingFace Models Hub

Foundation & Base Models

Fine-Tuned Specializations

IBM Granite for Enterprises

Vision-Language for Remote Sensing QA

Quick Access Collections

🛰️ Domain Imagery Gallery

🌍GeoSpatial Tech Stack

Transform

Analytics

GIS

AI

Python

Formats

Apps

Orchestrate

Processing

OLAP (Analytical)

Catalogs / Tables

Cloud Storage

🧭 Geospatial Fundamentals Cheat Sheet

🚀 HuggingFace Spaces & Interactive Apps

IBM-NASA Prithvi Demo Suite

Visualization & Analysis Tools

How to Use or Duplicate a Space

Frameworks & Deployment Options

Explore More Spaces

🎓 Courses to Learn Geospatial and Machine Learning

Free Courses

Paid Courses

Machine Learning for Remote Sensing (ML4RS) Workshop @ ICLR 2024 Microsoft Sponsor

Machine Learning for Remote Sensing (ML4RS) Workshop @ ICLR 2023

📁 Repository Contents [New additions are WIP]

🤝 Contributing

Show your support

📞 Contact

About

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Uh oh!

Contributors

Uh oh!

Languages

Packages