The Building Footprints pack provides access to information about the presence and shape of a wide variety of building types - from residential homes to large commercial and industrial buildings. Each building footprint is a polygon, which represents a simplified boundary for the perimeter of a building. It uses an additional machine learning model optimized for straight edges and sharp corners to reflect the regular geometric structures found in typical buildings.
In the Gen 1 to 5 AI Systems, the building footprint is identical to the roof outline, capturing any roof of a permanent structure such as a house, unit, commercial building, garage, large garden shed or carport that is designed to keep the weather out.
This explicitly excludes the top of a building, such as a rooftop basketball court or car park, which is designed for regular use by people. If a building is partly under construction, the completed part of the roof will be detected as building.
The vectorized building outlines exported from AI Parcels take the Roof AI Layer and perform post-processing to impose regular shapes, such as straight edges and 90 degree corners.
Gen 5 Building Footprints: A mix of residential buildings, commercial buildings, and garden sheds in Miami, Florida.
- Building - In Gen 1-5, the Roof layer is used to construct building footprints.
AI Feature API
(Gen 1 & 2 Data)
Standard Spec 1
"Building" feature class (vector polygons)
Gen 1 / 2 - Building Footprint presence; area estimate and polygon for each building footprint.
We introduced fidelity score on building footprint in gen4-lightning_bolt-1.0 for the first time. The fidelity score measures the quality of the vectorized building footprint polygon. One of the best uses for fidelity score is to programmatically identify buildings that require a different workflow or treatment due to CBD/downtown conditions or extensive tree overhang (e.g. falling back to a previous digitized outline, or flagging for manual inspection). Using a proprietary machine learning model that compares the final outline to the prediction rasters, we can accurately determine whether the polygon matches exceptionally well (1.0) or very poorly (0.0), or somewhere between.
The table below provides guidance on how to interpret a fidelity score.
The vast majority of buildings have a fidelity score of 0.9 or greater, indicating an excellent match.
An analysis of Gen 5 building footprints across more than 25,000 sq km of geographically diverse areas found an average fidelity score of 0.95.
|Fidelity scores between 0.7 and 0.9 will still be good enough for many applications, but may be degraded slightly due to challenging conditions such as extensive tree overhang, shadowing or architectural complexity.
|Scores between 0.5 and 0.7 may be useful in situations where the shape is less critical, but should be handled with care.
|The lowest scoring buildings often have low confidence as well, such as a heavily shadowed garden shed with overhanging branches. If the confidence is high but the fidelity is low, it is most likely a very large, complex commercial or industrial building that exhibits very complex architecture, or extensive parallax error (building lean due to height).
Fidelity scores in Omaha (left) and downtown New York (right). 0-1 scale mapped to red-yellow-green
Characteristics and Recommended Use
Commercial and Industrial Buildings
Tall Skyscrapers and complex buildings at the heart of many cities are very challenging, due to building lean (parallax) and complex multi-level structures where the reasonable extent of a footprint is ambiguous even to a human looking at nadir imagery. CBD/Downtown areas should be checked visually using MapBrowser, or programmatically using the Fidelity Score to confirm whether they meet the quality needs of your particular use case. Please discuss with your account manager if downtown areas are critical to your use case, as we have a range of options to assist you. Using the DSM directly is an alternative that may be useful for situations such as RF propagation modelling amongst tall buildings. You can find out about exporting DSM from MapBrowser here: Export 3D.
The "confidence" value for each building can be used in conjunction with the fidelity score to determine whether the building is captured well, or whether the model was uncertain. In many use cases, a threshold of ignoring buildings with a fidelity less than 0.4, or confidence less than 80% will result in a good balance between false positives and negatives.
Viewing the Building AI Layer in your area of interest can indicate whether the building footprints are likely to be of good quality in an area, or direct calculations using the DSM are a better option for you.
The building footprints are particularly good at identifying corners and edges of buildings that are obscured by trees (whether or not leaves are present). The AI Layer does a good job in the majority of these situations (except when the tree overhang is so large that it is difficult for a human to estimate how far underneath the roof extends), and enhancements in the Building Footprint vectorization process typically improve it further. We do not capture extreme overhang, where it is difficult for a human to estimate the extent of the building under the tree.
If the level of tree overhang is of interest, the actual shapes of overhanging trees and their areas are available in the Roof Characteristics AI Pack.
In an AI Offline Parcel export or via the AI Feature API Rollup endpoint, there are columns to indicate whether any buildings if at all are present, their total area, and the count. Checking for very low total area, a count of "0", or building present as "FALSE" indicates that it is likely a vacant lot. This is a good way to reliably identify vacant lots, or ones where the construction stage has not yet reached the adding of a roof.
The Construction and Surfaces AI Packs can be used if more granularity is required on the status of a vacant lot or it's state of construction.
Parcel Boundaries and Buildings
While many buildings lie entirely within a property boundary, several situations are common.
- In the AI Feature API, the entire building feature is returned if any part of the building intersects the Query AOI (usually a parcel boundary).
- The Rollup endpoint and AI Offline Parcel exports reference "clipped" and "unclipped" areas. Clipped areas are calculated on the fly to only include areas strictly intersecting the parcel boundary or Query AOI. Unclipped areas include the full area of the feature, which may extend over multiple parcels. The Rollup endpoint and AI Offline Parcel exports manage the following scenarios for you:
- Shifted Boundaries: Parcel boundaries are often shifted by up to tens of meters compared to ground. Small parts of nearby buildings may appear to protrude into the parcel, but should be ignored.
- Multi-parcel buildings: Many buildings legitimately span multiple parcels, such as condos, terrace houses, and townhouses. In this case, the "clipped" area of the building is most relevant.