Skip to main content

Canvas Gallery

Explore example Fused Canvases to see what's possible with Fused.

Overture Difference Heatmap
Overture Difference Heatmap

Visualize changes between Overture Maps releases

See Live demo →

Video Walkthrough

Canvas highlights

  • Real time processing of 2 hex datasets (one for each release)
  • Synchronized map viewport

Resources

Elevation & Temperature impact on crop types
Elevation & Temperature impact on crop types

Join datasets at different resolutions using H3 for dynamic crop analysis

See Live demo →

Video Walkthrough

Canvas highlights

  • Joining different datasets at different resolutions & projections using H3
  • Creating a dynamic interactive dashboard

Resources

Environmental Variables exploration
Environmental Variables exploration

Explore correlations between soil moisture, precipitation, temperature, and crop types

See Live demo →

Video Walkthrough

Canvas highlights

  • Soil moisture patterns
  • Precipitation data
  • Temperature variations
  • Crop type distributions

Resources

AIS Boat Exploration
AIS Boat Exploration

Processing AIS data to show trends in maritime traffic

See Live demo →

Video Walkthrough

Canvas highlights

  • Demo of the flexibility of ingested AIS data for complex visualizations
  • Creating multiple visualizations on the same dataset
  • Real-time maritime traffic analysis

Resources

AIS Dark Vessel Detection
AIS Dark Vessel Detection

Detect ships in Sentinel-1 radar images and flag the ones with no AIS position nearby

See Live demo →

Canvas highlights

  • Pick an example area or draw a box over US coastal water, then press Run to run the full pipeline
  • Loads Sentinel-1 GRD from Planetary Computer, masks land with ESA WorldCover and detects bright targets
  • Reads 270M geo-partitioned NOAA AIS pings and interpolates each vessel to the exact image time
  • Map of matched vs. unmatched detections, AIS tracks and a clickable list of candidates

Resources

AI Change Detection
AI Change Detection with Satellite Embeddings

Detect year-over-year land-cover change using 64-dimensional AEF pixel embeddings and cosine distance

See Live demo →

Canvas highlights

  • Draw any AOI on a satellite map, pick a year — all analysis nodes update automatically
  • Three synchronized views: raw embedding RGB, PCA(3) false-color, and cosine-distance change map
  • Change detection uses the full 64-D embedding space (not pixel subtraction), capturing semantic land-cover shifts
  • Reads only the two compared years from the global Zarr store — fast and cache-friendly

Resources

NDVI Time Series
Temporal Pixel Analysis — NDVI Time Series for Farm Fields

Sample Sentinel-2 NDVI at California field centroids over time and cluster farms by vegetation behavior

See Live demo →

Canvas highlights

  • Draw an AOI over California cropland, set a date range, and press Run — the full pipeline executes end-to-end
  • Queries CA field boundaries, builds a Sentinel-2 scene URL table, samples one pixel per field centroid
  • K-Means clusters fields by NDVI behavior (mean, variability) and renders a linked map + time-series chart
  • Hover a field on the map to highlight its NDVI curve in the chart, and vice versa

Resources

Monthly Composites
Monthly Median Composites — One Machine per Month

Fan out cloud-free Sentinel-2 monthly composites to one machine per calendar month and watch the grid fill in live

See Live demo →

Canvas highlights

  • Every calendar month in the date range runs on its own machine — 37 months finished in 6.2s wall time off 61.5s of work
  • SCL cloud masking plus a per-pixel median across each month's scenes, in NDVI or true-colour mode
  • Workers publish PNGs and status JSON to object storage instead of returning through the pipeline
  • The month grid polls storage every second, so cards fill in progressively as each machine finishes

Resources

Sentinel-2 Composite
Sentinel-2 Composites for Any Map View

Build a 10 m median composite of any three Sentinel-2 bands for the current map view on parallel machines, then download it as one COG

See Live demo →

Canvas highlights

  • Searches a STAC-GeoParquet mirror of Sentinel-2 C1 L2A with DuckDB and returns per-band COG URLs
  • Splits the view into 1024 px blocks and computes them on up to 200 machines at once with odc-stac and xarray (median, mean, min, max or first)
  • Each block is saved as a COG keyed by its settings, so re-runs and panning back reuse finished blocks
  • The blocks are merged into one downloadable COG (.tif), which also serves the map tiles at every zoom

Resources