unthinkable
Case study · A civil engineering consultancy, India · road surveys for NHAI

Highway planners get road survey reports quickly, without a person labelling every sign and barrier by hand.

The consultancy wanted to give the National Highways Authority of India fast, accurate survey reports for planning road work.

New Delhi, India

What changed

95%shorter throughput time (as reported)
88%average accuracy, expected to reach 93% (as reported)
90%less manual work (as reported)
What stood in the way

What stood in the way.

What stood in the way
  • A surveyor watched the camera feed from a car and labelled each road element by key press, so reports carried human errors and needed several reviews.
  • Objects, locations and distances were not identified accurately, and long-running barriers were hard to tell from single signs.
What changed
  • 95%shorter throughput time (as reported)
  • 88%average accuracy, expected to reach 93% (as reported)
  • 90%less manual work (as reported)
What we built

What we built.

  • An ML computer vision model on Keras and TensorFlow, trained on 1000 images for each category such as guard rails, beams, barriers and high tension wires.
  • Filtering of about 87,000 frames per video down to 15,000 relevant images and 1500 elements, in seconds.
  • Location from latitude, longitude and altitude in each frame, and automatic chainage that marks where a long element starts and stops.
  • Probability thresholds and multiprocessing to record likely objects faster, with the system installed on the consultancy's existing servers.
  • Keras
  • TensorFlow
  • Computer vision
  • Multiprocessing
What it shows

What this work shows about us.

Machine learning on the client's own data and servers

The model was trained and run where the survey video already lived, which kept costs down.

Built around the report the client delivers

Chainage and location for every element are what NHAI needs in the report, so the model was designed to produce them directly.

Want a similar machine learning road survey tool? Tell us what you want to change. Our AI assistant, Ask Unthinkable, helps you define your requirements, then connects you with the right team.

Start your plan ↗
If we built it today

If we built it today, with AI.

What we would add now, moment by moment.

Drafting the report

An agent would assemble the survey report from the detections, flag the elements the model was least sure of and send those to a surveyor to confirm. The surveyor still signs the report.

Planning the work

An agent would compare a new survey with the last one for the same stretch and list what has changed, for the planners to review.

Ask Unthinkable

Want this for your business?

Tell us what you run or what you're building. Our AI assistant, Ask Unthinkable, helps you define your requirements, then connects you with a senior AI engineer.

Send us a note instead

Required

Thanks. A senior AI engineer will reply to you.
u
Ask UnthinkableClarity first, then the right team
Your conversation

An AI agent. This prototype uses prepared answers and sends nothing.