Construction
Training data for construction AI, from the drawing to the site.
Takeoff drawings, 360° site progress photos and facade inspection imagery, annotated by reviewers trained on construction conventions.
From the drawing set to the finished building
Every part of the workflow where we can take work off your team.
Preconstruction
Estimating
On site
Finance
Inspection
We work with drawings, 360° photos, drone imagery and thermal.
What we label
Every class is written into a guide with your team before the first batch. Files come back as COCO JSON, YOLO, Pascal VOC, CSV, JSONL or your own schema.
Drawings and documents
Trade symbols
Bounding box and class
Outlet, fixture, valve, drain, door
Rooms and areas
Polygon
Room boundary, slab, ceiling zone
Linear runs
Polyline
Pipe, duct, cable tray, wall
Title blocks and schedules
OCR and key value
Sheet number, revision, scale, door schedule
Bills and invoices
OCR and line items
Vendor, cost code, quantity, amount
Site imagery
Progress states
Class and segmentation
Not started, framed, boarded, finished
Facade defects
Polygon and severity
Crack, spalling, corrosion, water ingress
Safety
Bounding box and keypoints
Helmet, vest, harness, person in zone
How we test the model
The tests that matter for construction models, and what you get back from each.
Takeoff count accuracy
Model counts checked symbol by symbol against a reviewed sheet.
You get backCount error by trade and symbol class
Extraction accuracy
Every extracted field on bills and invoices compared with the source document.
You get backAccuracy and errors by field
Reviewer agreement
Two reviewers label the same sheets blind, to show where the ontology is unclear.
You get backAgreement by class, disputed classes
Where construction teams start
Problems teams bring us, and the first piece of work we would scope.
Your takeoff model confuses symbols that look alike on different architects’ sheets.
What we doWe label across drawing styles, link each symbol to the sheet legend and log every ambiguous call.AnnotateExtracted cost codes reach billing before anyone checks them.
What we doA reviewer verifies each extracted field inside your pipeline, around the clock, until the model is ready on its own.Live reviewProgress photos pile up faster than anyone can compare them with the plan.
What we doPhotos labelled with a completion state per element, so the model learns what done looks like.AnnotateRelated work
A human verification layer for AI extracted fields on construction bills, invoices and site notes, running 24/7 until the model no longer needed it.
Live verification inside a production AI pipeline, under 2 minutes, for six months
Defect annotation on building imagery in CVAT, with a webhook loop between annotators and reviewers.
50,000 facade annotations a week for building condition assessments
Part identification and damage segmentation for insurance claims AI.
200,000 polygons for automated vehicle damage assessment
2,000 robot videos described every week by a team built in seven days
Bring us the construction data your model gets wrong.
Send a small batch and get it back done, free, usually within 24 to 48 hours. You judge the work before you pay for anything.