Nearmap AI contains a rich and continually growing library of content.  AI Layers  are the most basic unit of content; they feed into all of our AI data products, can be ordered as an offline data delivery, and in some cases can be visualised within MapBrowser. The AI Layer Glossary provides a list of all AI Layers produced by our models, along with definitions, examples, and links to any AI Packs that make them available.

What are AI Layers?

An AI Layer is a a single data layer that is processed on whole surveys of our imagery coverage. It represents a fairly direct output of our machine learning models, which is used in various ways in our post-processing pipelines. At present, they are available in two ways:

AI Packs: Some layers are included in an "AI Pack", and are enabled for visualisation in MapBrowser, and included as an input into a vectorised product that can be exported from MapBrowser. When visualised, these AI Layers form an overlay over our imagery to assist visual search.

Offline Delivery AI Layers can be ordered for offline delivery (downloadable files). AI Layers will often appear in the glossary before they have been made available through an AI Pack.

Note that where an AI Layer lists AI Packs as "Not yet available", it can still be ordered as an offline delivery. See  AI Offline Raster - Ordering and Delivery for details.

AI Layer relationships

An AI Layer may be related to a layer in a different AI Pack by virtue of sharing a common attribute. A relationship is indicated by the presence of the ParentID attribute, if no value is present then no relationships exist. 

For example, a Roof Condition layer identifies the condition of a roof as stained or damaged. This layer also needs information about the type of roof (tile, shingle, etc), which it obtains using the ParentID attribute that corresponds to the layer ID attribute in the Roof Characteristics pack.

IDs are not persistent between surveys and therefore relationships are only valid when layers have been created from the same survey date.

  • No labels