AI Offline Raster requires the most geospatial expertise to work with. While it technically contains more information than vector (each individual pixel, rather than a whole feature, has a confidence value), it does not include our enhanced vector processing, and metadata such as 3D derived attributes. Unless you really know you need raster, you are almost certainly better off with our vector products.

Why would I want a raster export of AI Layers?

There are two main reasons a raster export may be useful:

  • The AI Layer is not yet incorporated in an AI Pack as a vectorised product.
  • The raster data technically includes more information than the vector data; if you need to dig deeper than what the AI Pack vectorised data includes, then raster may be worth considering. The main consideration is if you want to choose your own threshold for vectorisation.

All AI Layers listed in the AI Layer Glossary are available for AI Layer Raster Export (Offline Delivery).

Who should consider offline delivery of raster AI Layers?

Geospatial raster data can be more complex to work with than vector products, or spreadsheet data. If you're not sure whether a raster AI Layer delivery is right for you, consider whether:

  • You have a defined area of interest (AOI) that could be drawn on a map. This delivery is not suited to a large number of small polygons scattered over a large area (such as an organisation that has interests in a few properties scattered throughout a state).
  • You have in-house GIS capability to deal with raster. If you are unfamiliar with concepts like thresholding a result to produce a binary map, have not used a tool such as ArcGIS or QGIS, or have not dealt with map projections such as EPSG:4326 and EPSG:3857, you may be better off using our AI Pack vectorised products.

Typical raster customers include local government departments.

What does the ordering process look like?

The process for ordering AI Layer Raster Export (Offline Delivery) is:

  1. Review the currently available AI Layers in the AI Layer Glossary.
  2. Discuss your needs with your Account Manager. Your Account manager may recommend a different delivery mode (Raster is typically more technically challenging to work with than the vector or spreadsheet-style data in the AI Packs).
  3. Submit an order, which requires you to provide:
    1. AOI: An Area of Interest (AOI), ideally as a geospatial polygon (e.g. geojson, kml or shapefile), or the name of a well defined government boundary.
    2. AI Layer Selection: A list of the layers you wish to purchase (including any composite layers - we can combine multiple layers into a single layer; combining Asphalt and Concrete into a single impervious surfaces layer is a common request).
    3. Date Selection: Any information about the date(s) you would like processed in your AOI (most recent, or a particular year, season, or date). It is possible to ask us to process an AI Layer on any of our imagery that has been published through MapBrowser or our imagery APIs.
    4. Map Projection: A map projection - by default, we produce EPSG:3857 tiles, however, we can convert these to a local projection of your choosing.
    5. Resolution: The raster AI Layers are produced natively at zoom level 21 in EPSG:3857. That means the highest meaningful resolution to request is approximately 7.5cm / pixel. Many customers prefer 30-50cm per pixel to reduce file sizes and make the processing easier to manage.
  4. Review the Order Summary: Attached to your quote will be a detailed order summary reflecting the above information, and what will be delivered. Please review it carefully to ensure it is correct.
  5. An email will be sent  with instructions on how to download  it once it is ready to go.

What do I need to know about what I receive? 

  • AI Raster offline is delivered as a link to a set of GeoTIFF files which are jpeg compressed, and split into tiles (depending on the size of the AOI and resolution). The link also contains a pdf describing the data, and a shapefile of the tile boundaries if the AOI has been split into multiple tiles.
  • Each pixel is represented by a 0-255 (uint8) value. 255 corresponds to an estimated 100% probability that the definition is present on that pixel (i.e. definitely present); 0 corresponds to 0% probability (i.e. definitely not present).
  • Most users will want to perform a threshold to create a binary layer; 127 (50% probability) is a good place to start, but it depends somewhat on the use case.
  • Where the AOI overlaps multiple imagery surveys, you will receive multiple sets of GeoTIFF files that overlap (we do not blend the surveys). There is a NODATA layer in the output to allow you to blend them as you see fit. This depends on the use case, and how much time has elapsed between the survey dates.
  • We can blend multiple different AI Layers into a single one for you (such as asphalt and concrete blended into an impervious surface layer), by statistically combining the probabilities on each layer.
  • You can choose a resolution up to 7.5cm/pixel, and a map projection.

What is the data format?


Can I see an example?

This GeoTIFF is an example of a single tile from within a delivery. This comes zipped with the other tiles from your AOI, and associated explanatory files:


AIRaster_8-MediumAndHighVegetation_09_49.tif