Home services and trades
Roof inspection photos and drone imagery: what makes them licensable
By SourceX Editorial · Reviewed by Noah Loul ·
Short answer
Roof inspection photos and drone imagery can be licensed for AI when five tests pass: the company owns or controls the images, each image carries capture metadata and a job link, labels record what it shows and what was decided, faces and addresses can be removed, and no carrier or vendor terms restrict reuse. Settle ownership first.
Key takeaways
- Photos taken by employees, contracted drone pilots, homeowners, carriers and measurement vendors carry different rights, so sort images by source first.
- A photo becomes data when it carries a capture time, a job link, the roof's material and the inspection stage.
- Labels that tie an image to a decision, such as repair, replace or supplement, separate a dataset from a photo library.
- Drone frames often capture neighbors' homes, vehicles and people, so wide shots need more privacy work than close-ups.
What makes a roof photo licensable?#
A roof photo is licensable when the company can show it controls the image, explain what the image shows and remove anything that identifies a person or property. Without all three, even a large library of clear hail and wind photos is hard to license.
Use five tests. Each answer is a yes, a no or a fixable gap, and the answers often differ between photos taken by your crews and imagery that came from someone else.
- Ownership: the company took the image or holds a written assignment or license that allows reuse.
- Metadata: capture time, job link, roof material, slope and inspection stage are recorded.
- Labels: the image is tied to a finding and a decision, not just stored in a folder.
- Privacy: faces, house numbers, plates and neighboring properties can be removed.
- Restrictions: no carrier, vendor or app terms limit reuse outside the job.
Who owns the image: crew, pilot, homeowner or vendor?#
Image ownership depends on who took the picture and what agreement covered the work. Photos employees take in the course of their job usually belong to the company, but drone pilots, subcontracted inspectors and vendors each need a closer look.
Fixing an ownership gap for future work is simple: add an assignment clause to pilot, inspector and canvasser agreements that covers photos, video and metadata. Fixing it for past imagery means asking for a written assignment, or leaving those images out until one is signed.
| Image source | Who usually holds rights | What to check |
|---|---|---|
| Employee photos in a job documentation app | The company, as employer | App terms on export rights and any vendor usage rights |
| Subcontracted inspectors or canvassers | Depends on the subcontract | An assignment or license clause covering photos |
| Contracted drone pilots | Often the pilot, unless rights were assigned | A written assignment of imagery rights, plus flight records |
| Third-party aerial measurement reports | The report vendor | Vendor terms; reuse beyond the job is usually restricted |
| Homeowner-submitted photos | The homeowner | Generally excluded |
| Carrier or adjuster photos | The carrier or adjusting firm | Generally excluded |
Metadata that turns photos into data#
Metadata turns a roof photo into a data point by recording when, where, on what roof and at what stage it was taken. Most job documentation apps capture time and location automatically; the rest has to come from the job record.
Provenance standards point the same way. The C2PA specification defines a Content Credential as a cryptographically bound structure that records an asset's provenance, with assertions about origin and modifications. A Content Credential shows whether that record is intact and signed, not whether the picture is accurate, so it complements job records rather than replacing them.
- Capture date and time, plus the date of the storm event or reported loss.
- Job ID linking the photo to the inspection, estimate and final invoice.
- Roof material, approximate age, slope and number of layers from inspection notes.
- Inspection stage: pre-loss, damage assessment, tear-off, decking, install, final.
- Device details and, for drones, altitude and camera angle where available.
Labels: what the image shows and what was decided#
Labels connect each image to a finding and a decision, which is the part an AI developer cannot recreate without your team's expertise. A hail bruise on a field shingle is useful; the same image tagged with the estimate line, the replacement decision and the carrier's approval is far more useful.
| Label | Example values | Where it usually comes from |
|---|---|---|
| Component | Field shingle, ridge cap, flashing, vent, skylight, gutter | Inspector tags or photo captions |
| Damage type | Hail bruise, wind crease, missing tab, granule loss, blistering, mechanical damage | Inspection notes |
| Severity | Cosmetic, functional, actively leaking | Inspection notes or report template |
| Decision | No action, repair, replace slope, full replacement | Estimate and signed scope |
| Outcome | Approved, denied, supplemented, reinspected | Carrier correspondence or job status |
Faces, addresses and other details to remove#
Faces, addresses and identifying property details are the main privacy work in a roof image set, and drone imagery carries the most of it. A wide drone frame can include neighbors' homes, people in yards, vehicles and house numbers that the job never needed.
Tools can handle a first pass, but check what each one actually detects. Presidio, an open-source toolkit for identifying and anonymizing personal data in text and images, includes a module for redacting personal information in images. A tool built to find text may catch a house number or a name on a mailbox and still miss a face, so faces, people and license plates often need a separate detection and blurring step. Presidio's own documentation warns that automated detection carries no guarantee of finding all sensitive information, so a person should review a sample of every batch, especially wide drone frames.
- House numbers, street signs and mailboxes.
- Faces of homeowners, neighbors and crew members.
- Vehicles and license plates in driveways and streets.
- Neighboring properties and backyards in wide drone frames.
- Interiors visible through windows or skylights.
- Precise GPS coordinates in file metadata, removed or coarsened.
Carrier, vendor and app restrictions#
Carrier, vendor and app terms are the restrictions that most often narrow a roof image set. Photos shared with a carrier under a preferred-contractor program may fall under that program's confidentiality terms, and measurement reports are typically licensed to you for the job, not for relicensing.
Read the terms of your photo documentation app as well. Check whether the company can export full-resolution originals with metadata, and whether the vendor claims any right to use your images. For drone work, keep pilot certification and flight records with the imagery, since a buyer may ask how the images were captured.
Illustrative: a storm-restoration roofer audits its photo library#
Illustrative: a fictional residential roofing company with a storm-restoration division stores crew photos in a job documentation app, uses a contracted drone pilot for steep roofs and orders aerial measurement reports for every estimate. The owner wants to know which images could be licensed.
The audit sorts images by source. Crew photos link to jobs, carry capture times and can be labeled from estimate lines, so they become the candidate set. Drone imagery is parked because the pilot agreement has no assignment clause, and the owner asks the pilot for a written assignment covering past and future flights. Measurement reports, homeowner photos and adjuster photos are excluded.
The company updates its subcontractor and pilot templates, adds a stage tag to the photo app and starts capturing the repair or replace decision on each photo set. The candidate scope is described in a fit check using metadata only.
How SourceX reviews roof imagery#
SourceX reviews roof imagery through the SourceX five-step transaction, Supply, Rights, Preparation, Approval and Delivery, with most of the work in Rights and Preparation. Each image source is matched to its rights holder, and only company-controlled images move forward.
The SourceX Evidence Packet then records provenance for each image source, the licensing rights relied on, the permitted use, a privacy record of what was blurred or removed, and the supplier's release authorization. Full-resolution archives do not need to move to SourceX: they stay in the company's own storage or ship on encrypted drives when delivery is approved.
Frequently asked questions
Are photos without job records still useful?
They are less useful. A photo with no job link, roof details or decision can show damage but not what was done about it. If job IDs can be recovered from file names, folders or dates, link them before deciding; otherwise those photos usually rank below linked sets.
Do drone images need different treatment from phone photos?
Yes. Drone frames are wider and more likely to include neighbors, people and vehicles, and drone files often hold precise location data. They also raise the pilot ownership question. Treat drone imagery as a separate source with its own rights check and privacy pass.
Should we strip GPS data from the files?
Precise coordinates are usually removed or coarsened before delivery, because they point to a specific home. A region, a storm date or a climate zone often gives a buyer what it needs without identifying the property. The right level depends on the permitted use in the license.
Can photos of commercial roofs be included?
Sometimes, but check the building owner's contract first. Commercial agreements more often include confidentiality terms, and some sites are security sensitive. Images of rooftop equipment, signage or access points may need removal even when the roof itself is fine to include.
Does our homeowner contract give us rights to reuse photos?
Many homeowner contracts mention documentation photos for the job or the insurance claim, which is narrower than reuse in a data license. The company usually owns photos its crews take, but how the contract and privacy notice describe their use still matters. Counsel can review the wording.
Sources
- The C2PA Explainer defines a Content Credential, also called a C2PA Manifest, as a cryptographically bound structure that records an asset's provenance, with assertions about origin, modifications and use of AI; Content Credentials do not judge whether provenance data is true. Source
- Presidio is an open-source, MIT-licensed SDK for PII identification and anonymization in text and images, with a module that redacts PII in images; its documentation warns that because it uses automated detection, there is no guarantee it will find all sensitive information. Source
Related resources
- QuestionDo AI labs buy images?
- InsightHow do I de-identify images and inspection photos for AI training?
- InsightHow do I de-identify inspection reports for AI training?
- InsightWhat compliance checks do food and beverage manufacturers need before licensing data to AI companies?
- SolutionData partnerships between businesses and AI developers
- IndustryBPO & contact centers data
See if your company qualifies
A short company assessment. No data uploads are needed.