Product Documentation

Generation 4 AI Content

Overview

Gen 4 is powered by a new raster model with the addition of eight new AI Layers for roof condition and construction. Gen 4 also improves existing layers already available in Gen 3. Open World vectorization has been expanded, enhancing some existing AI Packs and adding new ones. Gen 4 processing includes every survey from mid-2020 to present, and coverage now extends beyond Australia and the USA to include Canada and New Zealand.

Showcasing some of the new roof condition AI Layers

Version History

gen4-lightning_bolt

Version	Updates	New AI Packs	Enhanced AI Packs

`1.2`	New fields "extentMinX " , " extentMaxX " , " extentMinY " , " extentMaxY " added to the top level of the response. These represent the approximate extent of the AOI (either user supplied, or the result of geocoding). New fields " clippedAreaSqm " and " clippedAreaSqft " added to the top level of each feature in the response. These represent the area of each feature as if it were cropped to match the AOI. This upgrades all previously processed data from gen4-lightning_bolt-1.1 to gen4-lightning_bolt-1.2.
`1.1`	New fields "clippedAreaSqm" and "clippedAreaSqft" added to all feature classes behind AI Feature API. "clippedArea..." is the area of intersection between the provided Query AOI (usually a property boundary), and the feature. This makes it possible to look up the area of an object such as a building that falls within the AOI, and then calculate the proportion falling within or outside the AOI. This is important for determining which features to include and exclude. This pairs with the existing "unclippedArea..." attribute that gives the area of the entire feature independent of the Query AOI. This upgrades all previously processed data from `gen4-lightning_bolt-1.0` to `gen4-lightning_bolt-1.1`.
`1.0`	Uses the same raster model and vectorisation process, such that results will be identical to `gen4-building_storm.` Available on surveys processed from July 2021 onwards	Poles - Introduces a new feature in the data that captures the location of both power and light poles as a point feature with attached metadata.	Building Footprints introduces a new fidelity score for both building footprints and roof outlines along with confidence as a measure of quality. Roof Condition now includes a "low confidence roof condition" attribute for each of the five roof condition layers. The low confidence roof condition will describe a larger area, and a lower confidence than the standard one. Given roofs typically degrade gradually, this makes it easier to identify new roofs with no weathering at all (e.g. for tile and shingle staining), but has a much higher rate of false positives.

gen4-building_storm

Version	Updates	Enhanced AI Packs

2.2	Same as `gen4-lightning_bolt-1.2`. This upgrades all previously processed data from gen4-building_storm-2.1 to gen4-building_storm-2.2
2.1	Same as `gen4-lightning_bolt-1.1`.
2.0	A response includes all children of any feature that intersects with the query AOI, whether or not each child feature intersects with the parcel boundary. Previously, a response returned only features (children and parents) that intersected with the parcel. This upgrades all previously processed data from gen4-building_storm-1.x to gen4-building_storm-2.0	Construction - New AI Layers included in MapBrowser, and new feature classes in AI Feature API corresponding to: Building Under Construction Construction Crane Construction Vehicle
1.1	New fields "unclippedAreaSqm" and "unclippedAreaSqft" added to connected classes behind AI Feature API (vegetation and surfaces AI Packs). "areaSqm" and "areaSqft" will continue to give the area of the feature returned from the API. The "unclippedArea" includes the full area of the connected feature, which may be very large. This is to make it easier to determine if for example a property borders on a small stand of trees vs a large forest, or a pond vs an ocean. This upgrades all previously processed data from gen4-building_storm-1.0 to gen4-building_storm-1.1
1.0	First production data released as per the specifications below.

Coverage

The big new change to coverage with Gen 4 is adding New Zealand and Canada to our coverage footprints. Initial coverage for Gen 4 is every survey captured in all four capture countries from mid 2020 until launch in May 2021 (and ongoing).

The one key caveat to New Zealand and Canadian data is that AI Offline Parcel is not yet available (that is, raster and vector can be provided, but summary information trimmed to property boundaries and address information are not yet available).

New AI Packs

Roof Condition

The Roof Condition AI Pack includes these 5 layers visualized in MapBrowser, and AI Feature API + AI Offline vectorised data that intersects these with our building footprints. Roof Condition layers include:

Enhanced AI Packs

Surfaces	Water Body raster and vector has greatly improved performance.

Construction	Construction Site raster and vector has greatly improved performance. 3 other construction related layers are included in the Gen 4 model, but not yet added to the Construction Sites AI Pack. This will be done in future iterations.
Swimming Pool	Swimming Pool raster and vector has improved performance.
Solar Panels	Solar Panel raster and vector has improved performance.

New AI Layers

The Gen 4 model (gen4-building_storm-1.0) introduces a new raster model with eight new layers to the AI Layer Glossary. Some of these have not yet been included in generally available AI Packs from the first release of Gen 4, but will be subsequently added to existing and new AI Packs.