What hardware do I need to run a survey?

A consumer action camera and an RTK GNSS receiver on a standard vehicle. No sensor bar, laser profiler or dedicated survey vehicle is required. Gesix can also run the survey for you as a service.

How is the International Roughness Index calculated without a profiler?

IRI is estimated from the camera's own motion data and the vehicle's speed. It is intended for network-level screening and prioritisation. Where a contract specifies a particular profiler for acceptance testing, that requirement still applies.

How is accuracy checked?

Detections and positions are validated against terrestrial laser scanning (TLS) point clouds and RTK survey of the same road sections. Results are reported per defect class and per road condition, so your engineers can compare them with their own checks.

Does it replace a network survey vehicle?

It is designed to deliver comparable network-level inventory and condition data at a fraction of the capital cost. Whether it can substitute for a specific vehicle depends on the tender's technical specification, which we review with you during a pilot.

How does the output reach our asset-management system?

Results export as GeoJSON and CSV. They are structured to be ingested by our City Level Asset Management System or by any GIS or RAMS that accepts standard formats.

All Products

Centimetre-precision road surveys. Without the high CapEx.

GeoSurveyAI turns an action camera and an RTK GNSS receiver into a geo-referenced inventory of pavement distress, roadside assets and roughness, validated against terrestrial laser scanning (TLS) ground truth.

Schedule a Corridor Pilot Read the Technical Brief

IRC:82-2023
IRC:67
IRC:SP:16
RTK GNSS
TLS validated

Built forState PWDsHighway concessionaires (HAM / BOT)Smart CitiesUrban local bodies

The Pipeline

From a drive-through video to a queryable road inventory.

Four stages. Every output can be traced back to the frame it came from.

1
Capture
A consumer action camera records the corridor from an ordinary vehicle while an RTK GNSS receiver logs precise position. Video and position are synchronised frame by frame.
- Action camera
- RTK GNSS
- Any vehicle
2
Detect
AI finds roadside assets and outlines pavement distress such as cracking, potholes and rutting. Repeat sightings of the same defect are counted once.
- Roadside assets
- Pavement distress
3
Locate and measure
Each detection is placed on the map with absolute coordinates and measured for area, length and estimated depth. Roughness is estimated from the camera's own motion data.
- Coordinates
- Size
- IRI
4
Deliver
A structured GeoJSON and CSV inventory ready for asset-management systems, and an interactive web map with severity-coded distress.
- GeoJSON · CSV
- Web map
- Asset-register ready

Core Capabilities

Four pillars of an auditable road survey.

Pavement Distress Segmentation

Cracking, potholes and rutting are outlined in every frame and counted once across the drive, then measured in square metres, metres and centimetres.

IRC:82-2023
Severity scoring
Measured defects

Roadside Asset Classification

Signage, furniture and other roadside assets are detected, classified and placed on the map with absolute coordinates, trained on Indian road conditions.

IRC:67 signage
IRC:SP:16
Asset inventory

Low-Cost Roughness Estimate

The International Roughness Index is estimated from the camera's own motion data and vehicle speed, with no separate profiler sensor.

IRI
No extra hardware

TLS Ground-Truth Validation

Results are checked against millimetre-accurate terrestrial laser scanning (TLS) of the same road, so accuracy is measured and not assumed.

TLS ground truth
Independent checks
Position accuracy

The Interface

Camera evidence and map context, side by side.

Reviewers see the exact frame behind every detection next to its position and severity on the corridor map.

Illustrative interface. Layout and values are representative, not live survey data.

Technical Specifications

What is under the hood.

GeoSurveyAI technical specifications
Capture	Consumer action camera and RTK GNSS on a standard vehicle. No sensor bar or dedicated survey vehicle.
AI analysis	Multi-model detection and segmentation of roadside assets and pavement distress, with duplicate removal across frames.
Positioning	Every detection is placed on the map with absolute coordinates.
Roughness	International Roughness Index estimated from onboard motion data, intended for network-level screening.
Standards	IRC:82-2023 (pavement condition), IRC:67 (signage), IRC:SP:16.
Training data	Indian road imagery from public datasets and our own corridor surveys.
Outputs	GeoJSON and CSV inventories and an interactive severity-coded web map.
Validation	Terrestrial laser scanning (TLS) ground truth and RTK survey of the same road sections.
Deployment	Server, web or edge. Processing can run on your own infrastructure.

Verification

How accuracy is checked.

How results are checked before they are trusted.

01
Model selection
Candidate models are compared against TLS ground truth of the same road, using statistical tests, before one is adopted.
02
Training
Models are trained on Indian road imagery and combined, and duplicate sightings of the same asset are removed.
03
Position accuracy
Asset coordinates are measured against a manual RTK survey and compared with commercial road-survey platforms on the same sections.

Want the survey and the software together? Our terrestrial LiDAR team produces the ground-truth scans, and results flow straight into City Level Asset Management System.

Technical FAQ

Questions engineers ask first.

Book a 50 km corridor proof-of-concept.

Choose a representative stretch. We survey it, process it, and walk your engineers through every detection against your own ground truth.

Request a Pilot See City Asset Management

Centimetre-precision road surveys. Without the high CapEx.

From a drive-through video to a queryable road inventory.

Capture

Detect

Locate and measure

Deliver