Capgo | Brake Pad Identification App for Cyclists
Debbi Chandel
epairs moving.
The Challenge
Part numbers wear off, visual differences are tiny, and compatibility depends on brand, caliper and generation.
Catalog hunting took minutes per pad and led to returns and mis-picks.
Recognition had to work on busy benches with spotty connectivity.
What We Built
A guided camera flow with framing overlays and instant feedback, so blurry shots are retaken before they slow anyone down.
On-device recognition that returns an exact model with specs, supported brake systems and trusted purchase links.
Confidence indicators, clear error states and recovery paths so imperfect photos still lead somewhere useful.
Saved history and favourites for repeat orders, plus search and filters when no photo is available.
How We Did It
Build. One Flutter codebase for a consistent, responsive UI.
AI. TensorFlow Lite models running on-device for speed and privacy, with Azure vector search behind the catalog matching.
Backend. Firebase for auth, analytics and push, MongoDB for the catalog, compatibility mappings and model lineage.
Outcome
Decision time cut from minutes of catalog hunting to a few seconds.
Fewer returns and mis-picks for shops. Riders identify parts at home without specialist knowledge.
An error-prone task turned into a confident step in any repair workflow.
Like this project
Posted Oct 21, 2026
Snap a photo of a bicycle brake pad and get the exact model, compatibility notes and where to buy it. Flutter, on-device TensorFlow Lite, Firebase, MongoDB.
Likes
0
Views
1
