Mapping data services: map features, changes and places

Outsourced mapping data services give the detailed, rule-heavy work of building and maintaining map data to a trained team. Deepen AI's specialist team draws and attributes features from imagery, checks what changed, and reconciles places and addresses across sources, in your map editor or GIS, to your specification.

What the work is, and who buys it

Buyers are navigation and map data teams, delivery and mobility platforms that maintain their own maps, local search and place data teams, and HD map teams in automated driving. Their maps go stale all the time: lanes repainted, stores closed, new buildings, renamed streets, duplicate places from different feeds. Automated extraction handles much of the volume, but people still need to fix its output, verify changes and resolve conflicts between sources.

What our team does

• Draw and attribute lanes, signs, crosswalks, stop lines and points of interest from aerial or street-level imagery
• Read speed limits, turn restrictions and one-way rules from sign imagery, and attribute the road
• Verify map changes against fresh imagery, and flag what moved, opened or closed
• Reconcile addresses, boundaries and place data across feeds, and merge duplicate places
• Place pins on the correct building or entrance, following your pin rule
• Review map edits made by automated extraction or by another team

The team works in your map editor, QGIS, ArcGIS or an internal tool, on accounts you create. For turning recorded drives into tagged test scenarios, see scenario extraction.

How it runs

1. You share the specification and examples. Send your mapping specification, your source-priority rules, a few areas your team mapped well, and access to the imagery and the editor.
2. We agree scope and train. Our ops team reviews the work and sends the people, hours and rate in your pilot plan. Mappers train on your example areas, and we build a gold set of features and places with known-correct answers for you to confirm.
3. We deliver, with a named team lead and QA sampling. Mappers work through the live queue. The team lead answers specification questions and sends conflicts the rules do not settle to your contact.
4. You get a weekly report. Features added and changed, places merged, unresolved conflicts, QA sample results and specification questions.

Illustrative example: reconciling one place from two feeds

Illustrative only. The place, feeds and coordinates are invented.

• Input, feed A: "Lucia's Bakery" · 212 W 3rd St · pin 34.0491, -118.2512 · hours 7:00 to 15:00
• Input, feed B: "Lucias Bakery & Cafe" · 212 West Third Street · pin 34.0493, -118.2509 · hours 7:00 to 16:00
• Input, street-level imagery dated 2026-07: sign reads "Lucia's Bakery & Café", posted hours 7:00 to 16:00
• Output, match: same place. Merged, with the feed B ID kept as primary under your source rule.
• Output, name and hours: "Lucia's Bakery & Café", 7:00 to 16:00, both from the dated signage
• Output, address and pin: 212 W 3rd St in your format · pin moved to the entrance, per your pin rule
• Output, category: Bakery, with Cafe as secondary
• Who did what: one mapper reconciled the place. A QA reviewer re-checked the merge and the pin as part of the weekly sample and agreed.

How we keep map data reliable

Map quality depends on a clear specification and a clear rule for which source wins. The team lead turns each new question into a written ruling, and every mapper follows the same decision list. Gold features and places measure each mapper in training and in live work. QA reviewers re-check a sample of edits and merges against the imagery. They do not re-check every feature. Each decision records the date of the imagery it relied on, so your team can audit it later. The weekly report shows the acceptance rate on the sample and errors by type: geometry, attribute, merge and pin. See how we check quality.

Pricing scope

Mapping is quoted per project, above the standard rate. A place-reconciliation queue and a lane-attribution queue need different skills, so we price each setup separately. Tell us the feature types, the editor and a rough weekly volume in the pilot form. Your pilot plan then lists the people, hours and rate. The first two weeks run as a two-week prepaid pilot at that rate, with at least 20 hours per person per week, and include the team lead and QA sampling. From week three, invoices are weekly in arrears. Full terms are on the pricing page.

When to use the enterprise team instead

Mapping built from vehicle sensor data is enterprise work: HD map features extracted from lidar point clouds, multi-sensor recordings from your fleet, and calibration of the rigs that collect them. Deepen AI has delivered annotation and data services for Daimler Trucks, BMW, Bosch, Mercedes, Hexagon, Aptiv and Ford Otosan. The enterprise team scopes and prices each project. See enterprise AI data services.

Send your mapping specification and a sample area. We email you a pilot plan with a quote.

Request a pilot plan