Milk tests read by the phone camera, right on the farm
The producer photographs the rapid test and a computer vision model, running on the phone itself, returns the result in seconds. Every analysis is recorded with the farm, the sample source and the location.
- Client
- DairyTech
- Sector
- Agribusiness
- Line
- AI in production
- Period
- Oct. 2025 to Aug. 2026, with ongoing evolution
- Deliverable
- Android and iOS app
Technology
- Flutter
- On-device computer vision
- Native camera with cropping
- OpenStreetMap and GPS
- Token-based REST API
- Firebase Cloud Messaging
- Secure storage
A reading that depended on the eye of whoever was in the milking parlor
DairyTech works with rapid strip tests for milk: antibiotic residue in the tank and genetic tests for A2 milk and kappa-casein. Reading them depended on the eye of whoever was in the milking parlor, in poor light and with a phone in hand.
And what the lines mean changes from test to test. The result stayed on paper, with no link to the farm or the tank, and never reached the dairy that would receive that milk.
An app that reads the test and stores the result with photo and location
We built the entire product: the Android and iOS app, the backend and the computer vision model that reads the strip.
- Android and iOS appA guided 5-step flow: test type, farm, location, sample source and photo.
- Automatic strip readingThe app classifies the photo as one line, two lines or invalid image and translates it into the right result for each test.
- Camera with a frameAutomatic framing and cropping so every photo reaches the model in the same standard.
- Quality lockWithout enough confidence, the app does not guess: it asks for another photo and shows tips on what to improve.
- Farms, profiles and historyProducer, professional, dairy, cooperative and administrator, with history, map, filters and notifications.
The AI is the test reader
The producer drops milk on the strip, waits for the test time and points the phone. The model looks at the image and returns positive or negative, already translated for that type of test.
When the photo is too dark or out of frame, the AI does not guess. The approved result goes to the history with the original photo, for auditing by the dairy or the cooperative. The model can be updated without a new app release in the stores.
From the milking parlor to the dairy, everyone sees the same result
Three test types read by the same AI and five profiles in the dairy supply chain looking at the same result, with photo and location recorded.
Want AI like this in your operation?
Book 45 minutes on Teams with a specialist. We understand your context before proposing anything.
Next case: Wellu (app)