Overview
Gen 5 introduces a large number of new AI Layers, which brings the total number of raster layers to 80! This has resulted in a similar number of vectorized layers being available. The Gen 5 training dataset has increased by 2.6 times and has been focused on reducing uncertainty. We have also introduced the concept of status (Production, Alpha and Beta) to indicate the maturity of a layer against a tested use case.
The Gen 5 release coincides with the release of the Rollup API, which is an additional endpoint of the AI Feature API. The Rollup API returns a .CSV file with a set of attributes for the area requested. With Gen 5, the Rollup API produces more than 500 attribute columns such as the presence of water, vegetation, etc per polygon.
Pervious and Impervious Surfaces
Format and data changes
Confidence scores for absent features (with area of zero) will now be listed as "null" fields in JSON payloads (rather than "1" when absent). This will allow the use of the "null" value to filter absent attributes (rather than testing for e.g. an area of zero square meters).
Coverage
Gen 5 has been run on all surveys with an imagery publish date from July 2022 onwards, and other historical surveys on customer request. To assist with back testing and longitudinal analysis, a large historical catalogue of US cities has also been processed on an annual basis going back to 2015
Version history
gen5-urban_atlas-1.0
Version | Updates | New AI Packs | Enhanced AI Packs |
|---|
1.0 | This release introduces several improvements to processing speed and efficiency of the 3D building and roof characteristics (building height, building stories and roof pitch). Previously, an approximate match was made to a 3D mesh of the given survey ID to provide data for all surveys overlapping with recently produced 3D. With this release, 3D attributes are now only produced on the data from the 3D model exactly matching the survey ID requested, which improves the overall quality of the results significantly. This may require you to be more careful in your survey selection to obtain these attributes. For example, you can use the Coverage API to find a date with 3D resources attached.
Note: The 3D attributes (building height, building stories and roof pitch) will be available only in surveys with 3D capabilities.
This version is backwards compatible with previous Generation 5 releases.
Typically available on surveys processed from August 2023. | None | Building Characteristics Roof Characteristics |
Enhanced AI Packs
AI Pack: Building Characteristics
gen5-urban_atlas-1.0 computes 3D attributes (building height, building stories and roof pitch) only in surveys with 3D capabilities which are derived using the specific survey's data. This results in more accurate 3D attribute values.
gen5-tranquil_sea-1.0
Version | Updates | New AI Packs | Enhanced AI Packs |
|---|
1.0 | This is a minor release that includes: Improvements in building footprints accuracy on features greater than 8,000 square km.
Minor changes in area (typically less than 1%) for some feature classes due to subtle differences in raster predictions because of algorithmic optimization.
For Tree Overhang and Alpha and Beta layers, most cases where multiple polygons were provided in gen5-tranquil_pool will now be exploded into separate polygons. Each polygon has its own feature ID, confidence level and area rather than sharing these attributes as a group of polygons. The vector outline of each polygon remains the same.
This version is backwards compatible with gen5-tranquil_pool. Typically available on surveys processed from May 2023 onwards. | None | Building Footprints |
gen5-tranquil_pool-1.0
New AI Packs
Roof Objects
Roof Objects captures a range of objects and protrusions on top of a roof, that do not fall within other specific packs such as Solar Panels.
Debris
The debris pack captures several types of broken, disused and destroyed objects. These may be the result of a particular event (such as a hurricane), human behavior (putting rubbish in a dumpster), or degradation over time (abandoned cars, or washed up garbage).
Surface Permeability
Surface Permeability contains layers relating to whether water soaks into the ground, for storm water management work. Gen 5 introduces two new layers targeted specifically at impervious surface modelling. "Natural Pervious" captures all vegetated or unvegetated natural surfaces that are likely to absorb water. It covers a range of layers that have ambiguous boundaries - such as the distinction between healthy lawn, patchy lawn and dirt. "Hard Surface" captures a wider range of hard surfaces than just concrete and asphalt in a single layer - paving stones, brickwork etc. It also introduces a Natural Rock layer that captures rock larger than gravel that are cut in a natural style, such as at beaches and breakwaters, or rocky cliffs.
Surface Markings (Alpha)
Surface Markings focusses on painted markings on road surfaces, and providing a level of breakdown into key categories.
Enhanced AI Packs
Roof Condition
The Roof Condition pack was introduced in Gen 4 with 5x AI Layers. In Gen 5, those layers have all been improved further through adaptive labeling. Three new layers are also introduced (so there are 8x in total), targeting additional specific types of roof degradation.
Surfaces
Gen 5 introduces some surface layers that describe the usage, rather than the material. "Driveable Surfaces" largely include roads, but also other areas designed for vehicles to drive on - dirt roads, driveways, car parks, etc. The "Driveway" layer includes driveways and private roads made from any material. This can be used to identify entry/exit points to a property.
Swimming Pool
Gen 5 introduces a new feature to the existing swimming pool layer. The new layer identifies pools which require maintenance, likely due to lack of chlorination. The water is unsuitable for swimming, and can present the risk of mosquito borne disease.
Solar Panel
The Solar Panel pack had previously focused on photovoltaic panels only; the original layer has been renamed "Solar Panel (PV)", and a new "Solar Hot Water" layer has been added.
Vegetation
In Gen 3 and Gen 4, the vegetation pack included three classes of vegetation height. Gen 5 introduces a "Leaf Off Vegetation" layer, which identifies plants of any height without leaves, whether due to deciduous trees, plant health, or other reasons. Leaf off vegetation is useful both for understanding the state and health of vegetation, and understanding situations in which objects under trees (such as swimming pools) should be clearly visible. It is useful both at the level of an individual tree, or as an automated indication of the seasonality in a larger survey area capture.
Roof Characteristics
The shape of the roof is accessible via the Roof Characteristics pack (along with material, pitch and tree overhang). "Dormer Windows" is added as a new shape alongside Hip, Gable, Flat, Dutch Gable and Turret.
Gen 5 AI Layers
The Gen 5 status attribute results in three categories of layers:
- Production Layers: Semantically stable layers, which are suitable for general use and for which have been tested against a use case and determined to perform well.
- Beta Layers: Many of these layers are high quality, but are currently under customer testing for specific use cases.
- Alpha Layers: These layers are typically more experimental in nature and may not have been tested against a specific use case.
To evaluate a pack with Alpha or Beta layers in it, set the AI Feature API parameter "beta=true" or "alpha=true". These layers will also be available for offline delivery and are distinguished through the status attribute.
AI Pack | Layer | Status |
|---|
Roof Condition
| Missing Tiles and Shingles | Production |
| Permanent Repair (Patching) | Production |
| Zinc Staining | Production |
Surface Permeability | Natural Pervious | Production |
Hard Surface | Production |
Debris | Vegetation Debris | Production |
| Junk and Wreckage | Production |
| Discarded Tyres | Alpha |
Vegetation | Leaf-off Vegetation | Beta |
Swimming Pool | Unmaintained Swimming Pool | Beta |
Solar Panel | Solar Hot Water Panel | Beta |
Roof Characteristics | Dormer Windows | Beta |
Surfaces | Natural Rock | Alpha |
| Road (Driveable Surface) | Alpha |
| Driveway | Alpha |
Roof Objects | Chimneys | Alpha |
| Skylights | Alpha |
| A/C Units | Alpha |
| Solar PV | Production |
| Solar Hot Water Panel | Beta |
| TV Antenna | Alpha |
| Satellite Dish | Alpha |
Surface Markings (Alpha)
| Linear | Alpha |
Arrow | Alpha |
Block Symbol | Alpha |
Block Character | Alpha |
Painted Lane | Alpha |
| STOP Marking | Alpha |
| Pedestrian Crossings | Alpha |
|
