Safety-critical runway labeling for assisted-landing AI.

Challenge

Vision systems for assisted and autonomous landing must identify the runway with near-perfect consistency, day and night.

Airbus needed training data where every runway edge and light string was labeled to a safety-critical standard, with zero tolerance for drift between annotators.

Approach

io-ai stood up a dedicated team on a strict SOP. Daytime runways were labeled as four edge polylines (left, right, top and bottom, 2–8 vertices) forming a closed polygon with the fewest vertices needed for precise alignment.

Night scenes were labeled as runway light strings and green threshold lights, with vertices centered on the first and last bulb. Day and night data were separated before labeling to keep guidelines unambiguous.

Quality

Every task ran through io-ai's multi-level QA pipeline, with first-batch audits before scale-up and full rejection-reason metadata on every item, holding over 98% accuracy on rolling audits, with weekly status reporting.

Result

A consistent, audit-traceable runway dataset that fed Airbus's assisted-landing perception work, delivered to a safety-critical bar.

We also delivered cabin-interior panel segmentation: DADO, side-wall and wiring-harness polygons.