Structured
drone images
of real cities,
licensed for
model training

Aerial imagery for the Physical AI teams building world models and spatial intelligence.

25,000 km² (6.18 million acres) of U.S. and Canadian cities, captured since 2024, and growing.

  • Illustrated 3D tile of Miami, FL
    Miami, FL
  • Illustrated 3D tile of Vancouver, BC
    Vancouver, BC
  • Illustrated 3D tile of Chicago, IL
    Chicago, IL
  • Illustrated 3D tile of Halifax, NS
    Halifax, NS
  • Illustrated 3D tile of Phoenix, AZ
    Phoenix, AZ
  • Illustrated 3D tile of Toronto, ON
    Toronto, ON
  • Illustrated 3D tile of Seattle, WA
    Seattle, WA
  • Illustrated 3D tile of Ottawa, ON
    Ottawa, ON
  • Illustrated 3D tile of Calgary, AB
    Calgary, AB
  • Illustrated 3D tile of New York, NY
    New York, NY
  • Illustrated 3D tile of Montréal, QC
    Montréal, QC

Diverse terrain, sun angles, and seasonal variability

  • Illustrated 3D tile: Vancouver, BC, Pacific temperate rainforest
    Vancouver, BC
    Pacific temperate rainforest
  • Illustrated 3D tile: Edmonton, AB, Aspen parkland at the boreal edge
    Edmonton, AB
    Aspen parkland at the boreal edge
  • Illustrated 3D tile: Saskatoon, SK, Northern Great Plains
    Saskatoon, SK
    Northern Great Plains
  • Illustrated 3D tile: Phoenix, AZ, Sonoran Desert
    Phoenix, AZ
    Sonoran Desert
  • Illustrated 3D tile: Toronto, ON, Mixedwood plains
    Toronto, ON
    Mixedwood plains
  • Illustrated 3D tile: Houston, TX, Gulf Coast prairies and marshes
    Houston, TX
    Gulf Coast prairies and marshes
  • Illustrated 3D tile: Calgary, AB, Rocky Mountain foothills
    Calgary, AB
    Rocky Mountain foothills
  • Illustrated 3D tile: Highways, Interchanges, ramps and overpasses
    Highways
    Interchanges, ramps and overpasses
  • Illustrated 3D tile: Roads, Intersections, lane markings and crosswalks
    Roads
    Intersections, lane markings and crosswalks
  • Illustrated 3D tile: Residential parcels, Lots, roofs, driveways and yards
    Residential parcels
    Lots, roofs, driveways and yards
  • Illustrated 3D tile: Commercial parcels, Retail, parking and loading areas
    Commercial parcels
    Retail, parking and loading areas
  • Illustrated 3D tile: Tree species, Maple, oak, birch, cherry, spruce, pine and aspen
    Tree species
    Maple, oak, birch, cherry, spruce, pine and aspen

Cities shown are a sample of the archive. Ecoregion tiles show covered cities within their surrounding ecoregion. All scenes are illustrative.

Good aerial data your model can train on is hard to get

A stack of aerial image plates sits beside a tall licence document marked with a question mark, while a lowered barrier arm blocks the road out.?

Data provenance risk

Unclear image rights can stall a launch, or force you to retrain from scratch.

A city block is cut into a full grid of hexagons. A drone, tethered to its support van, captures one hexagon at a time while every other hexagon is still uncollected.

Data collection cost

Collecting the data yourself burns runway before training even starts.

Image prints with missing or mismatched labels lie scattered across a work surface beside sorting bins marked .jpg, .tif and a question mark..jpg.tif?

Data curation and preparation

Your best engineers spend time cleaning imagery instead of training models.

A car's street-level camera cone reaches only the side of a building, never its rooftop landing pad, and a patch of curb rendered in coarse blocky pixels loses the edge detail.H

Imagery diversity and resolution

Top-down airplane and satellite imagery with limited sun angles isn't enough to complement streetview datasets.

Value

Model Training is the solution

Model Training gives your team Spexi’s drone archive as one licensed, training-ready dataset. It answers each bottleneck with one contract, one transfer, one metadata standard and the high-resolution aerial view of every city block in its coverage.

A drone captures the building's rooftop landing pad straight down and the curb at an angle, both shown in green, alongside the car's existing street-level camera view. A chip reads 2.8 cm per pixel, and the curb's paint lines are crisp.H2.8 cm per pixel

The aerial view your model is missing.

Add the aerial view of every covered metro, from multiple angles and altitudes at 2.8 cm or sharper, to the ground-level data you already have.

A floating island cut in the exact shape of the continental United States, with layered soil along its edges, state borders on its green surface, and clusters of glowing green hexagons rising from covered cities from Seattle, San Francisco and Los Angeles to Denver, Houston, Chicago, New York and Miami. A label reads 32M+ images, 250 TB.32M+ images · 250 TB

Access the whole archive in one transfer.

License 32M+ images across 25,000 km² of U.S. and Canadian cities in one bulk transfer, and put the runway you would spend on a fleet into the model.

Sixteen drone image cards lie in a neat grid, each carrying the same metadata strip. A query panel above sets Spexi Image API parameters: capture type oblique, pitch from -60 to -30, heading from 90 to 180, and a capture date range, inside a bounding box for your area. Green dashed lines run from the query to the five cards that match, which lift off the grid with green outlines while the rest fade. A label reads Training mix, 5 of 16 match.Spexi Image API · Search imagesbboxyour areacapture_typeobliquepitch[-60, -30]heading[90, 180]captured_at[2025-03, 2025-09]Training mix5 of 16 match

Query the training mix. Don’t sort it.

Every image is captured to the same programmatic standard and carries the same metadata, so your team builds a training mix with a query instead of a sorting project.

The stack of aerial image plates now sits beside a single licence document marked with a check, and the barrier arm is raised so a car drives out on the open road.

One contract clears every image.

One licence covers every image, with rights to train models and deploy them commercially, so legal reviews a single contract instead of a pile of sources.

  • Close-up drone image of Winnipeg
    Winnipeg
  • Close-up drone image of Seattle
    Seattle
  • Close-up drone image of Portland
    Portland
  • Close-up drone image of San Francisco
    San Francisco
  • Close-up drone image of Los Angeles
    Los Angeles
  • Close-up drone image of San Diego
    San Diego
  • Close-up drone image of Las Vegas
    Las Vegas
  • Close-up drone image of Phoenix
    Phoenix
  • Close-up drone image of Salt Lake City
    Salt Lake City
  • Close-up drone image of Denver
    Denver
  • Close-up drone image of Dallas
    Dallas
  • Close-up drone image of Houston
    Houston
  • Close-up drone image of Austin
    Austin
  • Close-up drone image of Chicago
    Chicago
  • Close-up drone image of Detroit
    Detroit
  • Close-up drone image of Nashville
    Nashville
  • Close-up drone image of Atlanta
    Atlanta
  • Close-up drone image of Miami
    Miami
  • Close-up drone image of New York
    New York
  • Close-up drone image of Boston
    Boston

Aerial imagery is usually licensed for viewing. Training rights are a separate negotiation.

Model Training licenses the whole archive for training and commercial deployment, in one contract and one bulk transfer.

Typical aerial imagery licence

  • View in maps and reports
  • Train models on the imagery
  • Deploy trained models commercially
  • Bulk transfer of the full archive
  • Camera pose on every image

Spexi Model Training licence

  • View in maps and reports
  • Train models on the imagery
  • Deploy trained models commercially
  • Bulk transfer of the full archive
  • Camera pose on every image

Every image is a structured record

The Spexi pilot network captures cities in standardized, automated flight plans. Each image enters the archive with its footprint, camera pose and timestamps, so you can filter to the scenes your model needs most.

Spexi Image API · Search images

GET /developer-api/api/v1/image

p
lon,lat[,radiusMeters]
bbox
minLon,minLat,maxLon,maxLat
polygon
WKT, up to 1 km²
capture_type
panorama, orthomosaic, oblique
captured_at
[start, end]
heading
[0, 360), 0 = north
pitch
[-135, 45], -90 = nadir
altitude
metres above ground
limit
1 to 1000, default 100
cursor
next-page token
focused
default true
pixel_bounds
default false

→ GeoJSON features: footprint, camera pose, capture time

The result: query the exact imagery your model needs.

N

“All south-facing imagery in San Francisco, collected before 12 pm.”

bbox
-122.52, 37.70, -122.35, 37.83
heading
135 to 225 (south)
then keep
capturedAt before 12:00 local
  • South-facing drone image of San Francisco, Jul 11, 2024 · 10:46
    Jul 11, 2024 · 10:46
  • South-facing drone image of San Francisco, Jul 20, 2024 · 10:40
    Jul 20, 2024 · 10:40
  • South-facing drone image of San Francisco, Aug 23, 2026 · 11:04
    Aug 23, 2026 · 11:04
  • South-facing drone image of San Francisco, Jul 16, 2024 · 08:27
    Jul 16, 2024 · 08:27
  • South-facing drone image of San Francisco, Jul 9, 2024 · 07:08
    Jul 9, 2024 · 07:08
N

“-90° imagery of Phoenix, between 10 am and 2 pm.”

p
-112.07, 33.45, radius
pitch
-90 (nadir)
then keep
capturedAt 10:00 to 14:00 local
  • Straight-down drone image of Phoenix, Feb 10, 2025 · 10:15
    Feb 10, 2025 · 10:15
  • Straight-down drone image of Phoenix, Mar 22, 2026 · 11:09
    Mar 22, 2026 · 11:09
  • Straight-down drone image of Phoenix, Feb 8, 2025 · 11:29
    Feb 8, 2025 · 11:29
  • Straight-down drone image of Phoenix, Feb 19, 2026 · 10:42
    Feb 19, 2026 · 10:42
  • Straight-down drone image of Phoenix, Dec 26, 2025 · 11:55
    Dec 26, 2025 · 11:55
N

“North-facing imagery along I-5, between April and June 2025.”

polygon
I-5 corridor segments, ≤1 km² each
heading
315 to 45 (north)
captured_at
2025-04-01 to 2025-06-30
  • North-facing drone image of I-5, Portland · May 9, 2025
    Portland · May 9, 2025
  • North-facing drone image of I-5, Portland · May 9, 2025
    Portland · May 9, 2025
  • North-facing drone image of I-5, Los Angeles · Jun 15, 2025
    Los Angeles · Jun 15, 2025
  • North-facing drone image of I-5, Los Angeles · Jun 16, 2025
    Los Angeles · Jun 16, 2025
  • North-facing drone image of I-5, San Diego · Apr 5, 2025
    San Diego · Apr 5, 2025

Features and specs

Everything in a Living Coverage subscription, at a glance.

6 mo

Maximum image age

2.8 cm

Resolution (GSD)

25 acres

Per Spexigon

25 seats

Included per subscription

Three data products, all included

Also included

New captures in your area, no extra cost

Past captures already on file

Straight-down and angled views

Every capture verified before publishing

Unlimited API calls

Simple annual subscription

Add-on

On-demand capture

Need a site sooner? Task a new capture on your own timeline.

72 hrs

Turnaround in active coverage areas

Security, compliance, and data terms

Your data rights

Use imagery across your organization and in your own maps, reports, and applications for the life of your subscription.

Chain of custody

Every capture is timestamped, geo-located, and verified.

Freshness guarantee

If an area goes past six months, Spexi recaptures it and credits your account.

Access control

Start off with 25 user seats, with API keys managed by your admin.

Flight operations

Flown by vetted local pilots following Transport Canada and FAA rules.

Data residency

US customers' data is stored in the US, while Canadian customers' data is stored in Canada.

A drone shines a green capture beam onto a 3D model of a waterfront city.
Image API · Search ImagesGeoJSON Feature{"type": "Feature""geometry": { "type": "Polygon", … }"properties": {"id": "[image_id]""captureType": "oblique""capturedAt": "2026-09-14T15:42:08Z""processedAt": "2026-09-14T19:05:51Z""heading": 135"pitch": -45"roll": 0.4"altitude": 120"cameraPosition": { "type": "Point", … }"distance": 58.2"tag": "[tag]"}"links": [ … ]}Illustrative values · fields from the Search Images response
  1. 1. Captured
  2. 2. Verified
  3. 3. Published
  4. 4. Delivered

Frequently asked questions

Still have a question? Our team can walk you through coverage in your area.

What is Living Coverage?

It's an area-based entitlement that keeps imagery no more than six months old. It also includes the captures already on file and lets you request fresh imagery when you need it.

How often is my area refreshed?

Every area is recaptured at least every six months.

What if I need imagery sooner?

Request an on-demand capture. In active coverage areas, it's delivered within 72 hours of collection.

What happens if my imagery goes past six months?

Spexi recaptures the area and credits your account.

Can I subscribe to any area I want?

Subscriptions (referred to as entitlements) cover Spexi coverage zones, which include 100+ major U.S. and Canadian cities. Outside those zones, you can order one-off captures or buy existing imagery. Contact Sales to request a new area.

Can I buy a single image instead of a subscription?

Yes. If the imagery exists, buy it from the archive. If it doesn't, request a new capture.

Is Living Coverage right for construction monitoring?

Not on its own. Active job sites change faster than a six-month cycle. Use on-demand capture for milestone imagery, and Living Coverage to keep the surrounding area current.

What do I actually receive?

Static Images in nadir and oblique views, plus Standard or Aligned Orthomosaics as add-ons. Imagery is delivered through the Image API, Spexi Worldviewer, or ArcGIS.

What resolution is the imagery?

2.8 cm (1.1 inch) ground sample distance.

Does it work with ArcGIS?

Yes. Living Coverage imagery is available in ArcGIS through the Spexi Content Store for ArcGIS.

Who can use it in my organization?

Each subscription includes 25 seats, with unlimited API calls.

Can we share the imagery with other departments or the public?

You can use it across your organization and in your own maps and reports.

How is the imagery licensed?

Your subscription licenses the imagery for use across your organization, including in your own maps, reports, and applications, for the life of your subscription. Captures shared into your area by other customers come with the same rights. For public redistribution or other uses, contact Sales.

Is past imagery included?

Yes. Captures already on file for your area come with the subscription.

Where is imagery not captured?

Spexi doesn't capture over airports, correctional facilities, large bodies of water, or among tall downtown buildings, in line with flight regulations.

How is this different from scheduled aerial flyovers?

Aerial flyovers operate on a fixed cadence. Often, their maximum revisit is capped at every 6 months. Living Coverage is continuously collecting imagery to keep each image no older than 6 months the effective minimum refresh rate.

How much does it cost?

Living Coverage for major cities starts at $100 per sq km. Price depends on area size, add-ons, and current promotions.