Skip to main content
Public Preview

Get Started

Get started running RasterFlow in Wherobots.

Detect Objects with SAM3

Detect objects in imagery from natural language text prompts.

RasterFlow Datasets

Learn about built-in datasets and how to bring your own.

RasterFlow Models

Learn about built-in models and how to bring your own.

Build Multiscales

Add overview levels to a Zarr store so it can be visualized interactively.

RasterFlow Visualization

View mosaics, predictions, and vectorized results on a map.

Storage Integration

Read and write RasterFlow results directly in your own S3 bucket.

Run as a Job

Submit RasterFlow workflows as automated Job Runs.

RasterFlow Billing

Learn how RasterFlow usage is metered in RasterFlow Spatial Units.

Reference

Browse the RasterFlow API documentation
RasterFlow replaces the previous Raster Inference feature. For more information, see the changelog.
Wherobots RasterFlow is an image preparation and inference engine for large-scale raster processing and geospatial machine learning workflows. It enables you to build mosaics from multiple raster data sources, run inference with computer vision models, and vectorize results—all with a simple, high-level API.

​
Why choose RasterFlow?

RasterFlow is a managed workflow for geospatial raster processing tasks. It abstracts away the complexity of raster pipelines and distributed computing, allowing you to focus on your analysis rather than data engineering and infrastructure management. Key capabilities include:
Process raster data at massive scale with optimized chunking, sharding, and parallel processing
Abstract away the complexity with pre-configured datasets and models—or bring your own datasets and models.
Combine multiple raster datasets into unified mosaics that are ready for further analysis
Apply machine learning models to massive raster datasets at scale
Convert raster predictions into vector geometries for spatial analysis
Add overview levels to mosaics and model outputs so they can be viewed interactively on a map
Build flexible pipelines with your own data, models, and processing parameters
Work with GeoTIFF, Zarr, and GeoParquet formats

​
Key concepts

The following concepts are fundamental to understanding how RasterFlow works:
Mosaics are spatially-aligned raster datasets stored in Zarr format. RasterFlow can build mosaics from aerial and satellite imagery to prepare them for model inference. This process can combine one or more datasets across an Area of Interest (AOI) and a temporal dimension to create a single seamless input mosaic for inference.RasterFlow has workflows for building mosaics from the built-in datasets or your own imagery.
Run computer vision models on raster data at scale:
  • Semantic Segmentation: Classify each pixel (e.g., land cover mapping)
  • Regression: Predict continuous values (e.g., canopy height estimation in meters)
  • Patch-based Processing: Handle large mosaics by dividing into manageable patches
Convert raster predictions to vector geometries:
  • Threshold-based: Binarize continuous predictions
  • Polygonization: Create polygon features from classified pixels
  • Coordinate transformation: Reproject to desired CRS (e.g., WGS84)
RasterFlow writes mosaics and model outputs as Zarr stores at a single native resolution, which suits analysis but makes interactive viewing slow. Building multiscales reads an existing store and writes a new one that adds downsampled overview levels—an image pyramid—plus histogram statistics, so a map client can stream coarse tiles when you are zoomed out and full-resolution pixels when you zoom in. The source store is left unchanged.Build Multiscales with RasterFlow

​
Get started

You can run RasterFlow from Wherobots Cloud or from your own editor with the Wherobots VS Code Extension. Both paths execute against a Wherobots runtime, so the RasterFlow code is identical. RasterFlow Model Hub
1

Go to the Model Hub

Login to Wherobots Cloud and try out one of the built-in models by running a RasterFlow notebook in the Model Hub.
2

Click Run Model in Notebook

Select your desired notebook and click Run model in Notebook to provision a runtime.RasterFlow Model HubOnce the runtime is provisioned, a JupyterLab interface will appear with your chosen example notebook.
Clicking Run Model in Notebook from the Model Hub will open the solutions notebook with a Tiny runtime.
3

Run each cell

Click Run > Run all cells in JupyterLab.
RasterFlow manages its own compute resources for raster processing, so the Wherobots Runtime size you select does not affect RasterFlow workflow performance. You should generally use the Micro runtime for RasterFlow workloads to minimize cost.The only exception is if you plan to also perform vector processing with WherobotsDB (e.g., using SedonaContext for spatial SQL queries) in the same job or notebook session. In that case, choose a runtime size appropriate for your WherobotsDB workload — the RasterFlow portions of the workflow will still be unaffected by the runtime size selection.

​
Included example notebooks

RasterFlow comes pre-configured with five open-source models:
  • Meta SAM3 for text-prompted object detection
  • Fields of the World for agricultural field boundaries
  • Tile2Net for sidewalks and crosswalks
  • Meta CHM v1 for canopy height
  • ChesapeakeRSC for rural roads.

​
Run example notebooks

The following pre-configured model solutions notebooks are currently loaded in Wherobots Cloud to showcase the core features of RasterFlow:
sam3-thumb

Text-Prompted Object Detection

Detect objects in aerial imagery from natural language prompts and return georeferenced geometries using the Meta SAM3 model
ftw-thumb

Agricultural Field Mapping

Detect field boundaries from Sentinel-2 imagery and segment crop fields across regions using the Fields of the World model
tile2net-thumb

Urban Infrastructure Detection

Identify sidewalks, crosswalks, and pedestrian pathways from high-resolution aerial imagery using the Tile2Net model
chm-thumb

Canopy Height Estimation

Predict tree canopy heights from aerial imagery to monitor forest health and vegetation structure using the Meta CHM v1 model
chesapeakersc-thumb

Rural Road Detection

Identify roads in rural environments and map road networks using the ChesapeakeRSC model
The examples below demonstrate real-world applications that you can adapt to your own data and use cases. Each notebook provides a complete, working implementation to help you get started quickly.

​
RasterFlow use cases

  • Detect field boundaries from Sentinel-2 imagery
  • Segment crop fields across counties/regions
  • Convert raster predictions to vector geometries
Example application: Map all agricultural fields in Haskell County, Kansas using Sentinel-2 imagery and the Fields of the World modelTry the Fields of the World notebook
  • Identify sidewalks, crosswalks, and pedestrian pathways
  • Generate detailed maps for urban planning
  • Analyze accessibility from high-resolution aerial imagery
Example application: Detect and map sidewalk networks in College Park, Maryland using 30cm NAIP imagery with the Tile2Net modelTry the Tile2Net notebook
  • Predict tree canopy heights from aerial imagery
  • Monitor forest health and vegetation structure
  • Support conservation and urban forestry initiatives
Example application: Estimate tree heights across Nashua, NH using 60cm NAIP imagery with the Meta CHM v1 modelTry the Meta CHM v1 notebook
  • Identify roads, especially in rural environments
  • Map road networks to support routing and navigation
  • Detect road network changes to keep maps up to date
Example application: Detect roads in Maryland using 1m NAIP imagery with the ChesapeakeRSC modelTry the ChesapeakeRSC notebook
  • Detect objects from a natural language prompt
  • Return georeferenced bounding boxes or polygons with confidence scores
  • Skip vectorization, since output is already vector
Example application: Detect roofs in College Park, Maryland from 30cm NAIP imagery using the SAM3 modelTry the SAM3 notebook
  • Run a model over bi-temporal imagery stacked along the channel dimension
  • Compare two seasons or years in a single inference pass
  • Identify where predictions changed between time periods
Example application: Compare planting and harvest season imagery for Haskell County, Kansas using the Fields of the World modelView the Change Detection notebook
  • Build cloud-free median composites from Sentinel-2
  • Use built-in planting and harvest season datasets, or a custom date window
  • Return all 12 L2A spectral bands at 10m
Example application: Build a seasonal Sentinel-2 mosaic over Haskell County, Kansas as the input for agricultural inferenceView the Sentinel-2 mosaic notebook
  • Build mosaics from built-in NAIP datasets
  • Reproject source imagery onto your own output grid
  • Produce 30cm and 60cm analysis-ready rasters
Example application: Build a 30cm NAIP mosaic over College Park, Maryland ready for model inferenceView the NAIP mosaic notebook
  • Query a STAC catalog to discover imagery for an area of interest
  • Create a GDAL Raster Tile Index (GTI) from STAC items
  • Build a seamless mosaic from your own tiles with build_gti_mosaics
Example application: Mosaic NAIP imagery discovered through STAC over Montgomery County, Maryland, then optionally run road detectionView the Bring Your Own Rasters notebook
  • Export a custom PyTorch model to a RasterFlow-compatible format
  • Register the model and run it as an inference task
  • Apply your own architecture at scale over a mosaic
Example application: Export a PyTorch model and run it over RasterFlow imagery alongside the built-in modelsView the Bring Your Own Model notebook

​
Next steps

​
API reference

For detailed API documentation, see: