Skip to main content

Diagnostics

Yolk Color

In development · not yet validated

Photograph a cracked egg on a light plate, and record a colour result against the flock that laid it — in the same place as its production data, feed programme and reports.

Where this actually stands

The capture, upload, review, history and reporting workflow is built. The scoring method behind it is not yet validated. Devenish has not finalised the reference scale, the calibration procedure against a physical standard, or the labelled validation set the method must be tested against.

Until that work is complete, the platform does not present a yolk colour score as an approved measurement, and this feature stays disabled in production. We would rather ship a feature late than have a producer make a feed decision on a number we cannot stand behind.

The workflow

What the module does today

Everything except the scientific interpretation is real, tested and audited.

  1. Guided capture

    Camera overlay, lighting and focus prompts, and distance guidance in the mobile app. Photos can also be uploaded from a phone gallery or a computer.

  2. Image-quality checks

    Blur, exposure and glare are assessed before anything else happens. A poor image is rejected with a reason rather than force-scored.

  3. Context, not an orphan photo

    Every capture is attached to a customer, farm, barn and flock, so a result means something later.

  4. Processing status

    Queued, processing, complete or failed — visible while it happens, retried safely if a connection drops.

  5. Nutritionist review

    A result can be reviewed, corrected and commented on before it counts, with the correction recorded.

  6. History and reporting

    Results build a trend for the flock and can be included in consultation reports.

Validation

What has to be true before this is switched on

  • An approved reference scale, agreed with the Devenish nutrition team.
  • A calibration procedure measured against a physical reference standard rather than a photograph of one.
  • A labelled validation set covering the lighting and handling conditions of a real packing room.
  • An accuracy expectation that Devenish has reviewed and accepted.
  • A defined workflow for what happens when the system is not confident.

Each of these is tracked as an open item in the project's own gap register rather than left as an intention. The diagnostic framework is shared, so the same bar applies to FeedScan and to any future module.

Interested in the diagnostics pilot?

We are looking for operations willing to help validate the method under real conditions.

Talk to us about the pilot