products
Runway Surface Scan by Drone at a Known Resolution
Runway and taxiway pavement scan by drone at a known ground resolution, in active development. The platform already plans flights and georeferences…
AUG 03 2026
11 MIN
ROAD PAVEMENT · MUNICIPAL NETWORKS
In development
The capture is built, crack detection is on the way
The service is in active development. The platform already plans the flights and captures drone and ground imagery over a window of one to two hours, georeferenced and retained. Crack detection is close to a first MVP and finds cracks only.
1–2 h
capture window for the agreed area
cracks
the only defect type in development
none
road closure or lane booking beyond that window
retained
georeferenced imagery, ready for a newer model
FIT
Who this is for
A road department knows which streets are bad. What it usually cannot produce on demand is the evidence, in the same form, for the whole network, updated this year.
For
Municipal road departments that own a network and rebuild their condition picture by driving it and writing notes.
City GIS teams that already hold a road centreline layer and want a repeatable imagery capture they can refresh each year.
Regional road authorities preparing a maintenance budget who want a consistent imagery record of the network before the budget round.
Not for
Networks that need a formal ASTM D6433 survey signed by a certified inspector for a funding programme.
Structural or bearing capacity questions, which need coring or deflection testing rather than surface imagery.
Single-street complaint investigations, where sending a crew to look is faster and cheaper than a survey.
JUDGED ON DELIVERY
Outcomes and acceptance criteria
The survey exists to replace an opinion with a record. Three outcomes carry that, and each can be checked on the day it is delivered.
01
Every agreed street imaged
The agreed streets come back as georeferenced imagery from drone and ground capture, taken in a window of one to two hours for the agreed area.
Acceptance
Pick any street in the agreed area, and the imagery covering it is there with its position.
02
Cracks found, as it matures
Crack detection on the imagery is in development, close to a first MVP. It finds cracks as one class and nothing else, and no ASTM D5340 codes are claimed for roads.
Acceptance
The pilot scope says what has been run on your imagery and what has not.
03
Imagery a newer model can be run on
Imagery is retained, so a survey can be re-run on a newer model. The re-run is manual today.
Acceptance
The retained imagery is there to re-examine on request.
A condition rating, a priority list and a 3D model of the surface are not part of this service. Comparing one year with the next is done by your engineer on the retained, georeferenced imagery.
What the survey does not produce is a repair design, a bill of quantities or a cost estimate. Those belong to your designer. A condition survey that quietly becomes a works specification is one nobody can hold to account.
SCOPE AND DELIVERABLES
What is inside the survey
Each deliverable with what it contains and who owns it, including the parts that are deliberately outside the scope.
11 parameters
Crack detection, in development
Deliverable
What it contains
Owner
01
Georeferenced imagery
Drone and ground imagery of the agreed streets, retained so that it can be re-examined.
Owner
TarmacView
02
Crack detection
In development, close to a first MVP. Cracks only, as one class, run on a separate GPU machine. No distress codes are claimed for roads.
Owner
TarmacView
03
Prototype output
In development. A section and sample-unit inventory as GeoJSON plus a JSON and Markdown report, keyed by section and sample-unit ID.
Owner
TarmacView
04
Re-run on a newer model
Retained imagery lets a survey be re-run on a newer model. The re-run is manual today.
Owner
TarmacView
05
Capture window
One to two hours for the agreed area, drone and ground imagery, no closure booked beyond that window.
Owner
TarmacView, with the road authority
06
Flight permissions
Arranged before the date where local rules require a permit for the drone flight.
Owner
TarmacView
07
Section map
The network split into sections, agreed before capture and reused on later surveys.
Owner
Municipality, from its own centreline data
08
Condition rating, priority list and 3D model
Not included. The service does not produce a rating, a ranked list or a surface model.
Owner
Excluded
09
Repair design and costing
Not included. The survey records the roads, it does not specify works.
Owner
Excluded
10
Structural evaluation
Not included. Bearing capacity needs coring or deflection testing.
Owner
Excluded
11
Traffic and safety assets
Not included. Signs, markings and drainage assets are a separate inventory task.
Owner
Excluded
Crack detection is in development and is not an ASTM D6433 survey. No ASTM D5340 codes are claimed for roads.
DELIVERY · 6 STEPS
From centreline layer to georeferenced imagery
01
Pilot conversation
You send the streets you have in mind and, if you have one, your centreline layer. We agree the section split, how results are handed over and the capture windows. Output is a written pilot scope. Without an agreed section split the pilot does not start, because results against boundaries nobody accepts cannot be defended.
02
Permissions and planning
Flight permissions are arranged where local rules need them, and the capture is planned from the road geometry and the camera optics so that coverage is computed rather than assumed. Output is a validated plan and confirmed dates.
03
Capture
Drone and ground imagery over one to two hours for the agreed area, with traffic moving through it. Larger networks are captured across several windows. If weather stops a window, that window is reflown rather than the plan rebuilt.
04
Crack detection, in development
Imagery is built into an orthomosaic and run through the crack model on a separate GPU machine. This step is close to a first MVP and finds cracks only. No distress codes are claimed for roads.
05
Section inventory
In the prototype output, detected cracks are grouped by agreed section and sample unit as GeoJSON, with a JSON and Markdown report. No rating per section and no priority ranking is produced.
06
Delivery
The imagery and the outputs agreed in the pilot scope land together, and we walk your engineer through them. Anything disputed is re-examined against the retained imagery.
EVIDENCE AND CAPABILITY
How the detection works
The part of this service worth understanding before buying it is the detector, because it is the part still in development.
Images are passed through self-supervised vision transformer backbones, DINOv2 and DINOv3, which are used for different purposes. They were trained without labels on a very large image corpus, and it turns every small patch of an image into a high-dimensional embedding that carries texture and structure rather than a category name. A lightweight head sits on top of those embeddings and does the one job in scope, segmenting a crack from the surface around it. Finding a pothole , telling longitudinal cracking apart from a sealed joint and grading alligator cracking are not in scope. The method page is DINOv3 for infrastructure surface analysis , which goes into the newer of the two backbones properly.
What matters operationally is not the architecture but the consequence. Frozen backbones with a trained head apply the same decision boundary to every metre of every survey. A crew driving a network applies a threshold that drifts with fatigue, weather and who is in the vehicle. The machine is not more perceptive than a good engineer standing on the road. It is more consistent than the same engineer on the twentieth kilometre, and consistency is what a time series needs.
No rating is built on those detections. A Pavement Condition Index is a statement about a defined section, and ASTM D6433 asks an inspector to walk sample units and read deduct curves by hand. TarmacView does not produce a PCI or a proxy for one. Where a funding programme names the standard, the formal survey is what satisfies it.
The retained imagery is the record. It lets a survey be re-run on a newer model, which is manual today, and lets your engineer compare one year with the next. A pavement management system can load GeoJSON, but no direct export to one is built, and the hand-over format is agreed before capture.
PRICING
Not offered yet
This service is in active development and is not offered yet, so there is no price, rate or quote. Four things shape a pilot instead.
EUR
not offered yet
excl. VAT
EU
Checked 2026-09-04
Cost driver 01
Length and width of road
Capture time and processing both follow it.
Cost driver 02
How finely the network is sectioned
The same imagery produces a different output at ten sections than at a hundred.
Cost driver 03
Ground imagery alongside the drone
Needed or not depending on tree cover, parked vehicles and how narrow the streets are.
Cost driver 04
Travel
Travel to the survey area.
The detection model improves on survey imagery. Whether your imagery may be used for that is written into the agreement rather than assumed, and the capture you receive is the same either way. To follow the development or try it early, talk to us
.
EACH WITH AN OWNER
Risks and dependencies
Four things can stop a survey or weaken its result, and each has an owner.
Owner · WEATHER
Weather and surface state
Standing water, snow and heavy leaf cover hide the surface from a camera. Results are best on a calm, sunny day, because high wind blurs images and low light forces a high ISO. The affected window is recaptured rather than analysed through the obstruction, and this is the most common reason a date moves.
Owner · BOTH, AT SCOPING
Occlusion
Parked cars, overhanging trees and narrow streets with tall buildings block a view from above. This is why ground imagery is in scope for some networks and not others, and it is decided at scoping rather than discovered on the day.
Owner · TARMACVIEW
Flight permissions
Urban drone flight is regulated, and a permit that does not arrive is a date that does not happen. That work sits with us, but it depends on the authority granting it.
Owner · BOTH, AT SCOPING
Section agreement
Results describe a defined stretch of road. If section boundaries move between surveys, the comparison breaks. Boundaries are frozen at scoping, and changing them later is a decision with a cost rather than an edit.
Accountability sits with TarmacView for capture, and for crack detection once it is delivered, and with the municipality for the section map, for access and for what is done with the results. Nothing here is a compliance statement, and whether a given condition record satisfies a particular funding programme is a question for that programme.
Frequently Asked Questions
- What exactly arrives at the end of a survey?
- The service is in active development and not offered yet. The platform already plans the flights and captures georeferenced drone and ground imagery of the agreed streets, retained so it can be re-examined. Crack detection is close to a first MVP and will find cracks only. Its prototype output is a section and sample-unit inventory as GeoJSON plus a JSON and Markdown report. A condition rating, a priority list and a 3D model are not part of the service.
- Do you need to close the road?
- No closure is booked in advance. Capture runs from drone and ground imagery over a window of one to two hours for the area agreed, and traffic keeps moving through it. Where local rules require a permit for the drone flight, that permit is arranged before the date rather than on the day. Larger networks are captured across several windows rather than in one longer closure.
- How does the crack detection actually work?
- Imagery is built into an orthomosaic and passed through a vision model with one class, crack or not crack, run on a separate GPU machine. It is in development, close to a first MVP. The backbones are self-supervised vision transformers, DINOv2 and DINOv3, used for different purposes, which produce an embedding for every small patch of the image, and a lightweight head on top of those embeddings segments the cracks. The practical consequence is that the detector applies the same threshold everywhere, on every survey, which is the property a human crew driving a network cannot offer.
- Do you report a Pavement Condition Index?
- No. TarmacView reports neither a Pavement Condition Index nor a proxy for one, and no condition rating per section. The idea behind the index is that detected distress of a given type, severity and density reduces a score from 100 downwards, and a formal survey has an inspector walk sample units and read deduct curves. Only crack detection is in development for roads, and no ASTM D5340 codes are claimed for roads. Where a programme requires a formal survey to a named standard, a formal survey is needed.
- Is our imagery used to train the detection model?
- Only if you agree to it. The detection model is in development and survey imagery is what improves it, which is worth something to us. Whether your imagery may be used that way is written into the agreement rather than assumed. If you do not want it used, say so at scoping and the capture is done on that basis.
- Can the output go into our existing systems?
- Not directly yet. The prototype output is a section and sample-unit inventory as GeoJSON plus a JSON and Markdown report, and a municipal GIS can open GeoJSON. No import or export for a particular pavement management system is built, so how results are loaded into yours is agreed before capture.
References
- astm-d6433
- ASTM D6433-24, Standard Practice for Roads and Parking Lots Pavement Condition Index Surveys ASTM International
- dinov3
- DINOv3 (arXiv:2508.10104) Meta AI Research and Inria
- fhwa-ltpp-dim
- Distress Identification Manual for the Long-Term Pavement Performance Program, Fifth Revised Edition (FHWA-HRT-13-092) Federal Highway Administration
- faa-ac-150-5380-6
- AC 150/5380-6, Guidelines and Procedures for Maintenance of Airport Pavements Federal Aviation Administration
Reviewed by Štefan Moravík · 2026-09-04
Keep reading
03 PAGESLearn more
products
Airfield Pavement Condition Index (PCI) by Drone
A drone survey to an ASTM D5340 distress inventory and a PCI for every sample unit of the airfield, with every crack measured, rated and shown on one…
SEP 04 2026
8 MIN
solutions
US Military Airfield Pavement PCI Surveys by Drone
A PCI for every sample unit of a US military airfield, from drone imagery and the D5340 definitions. Early access, in development, with the limits…
OCT 18 2026
13 MIN
Get Started
Talk to us about the development
This service is in active development and is not offered yet. If you want to follow it or try it early on your own roads, tell us the streets you have in mind and we will talk it through. Nothing is binding.