What is an unverified egg, and when does the system not grade it?

What is an unverified egg?

An unverified egg is one the system does not grade, because the image does not contain enough information for a reliable class. It is not clean, dirty or cracked. It is an egg that needs a human look before any decision.

In Vigiovo's class list, unverified sits alongside clean, dirty and cracked. It is not a fifth defect: it is the system declining to judge an egg the image does not show in full. The term is defined in the glossary.

Which rules mark an egg as unverified?

An egg is marked unverified when it fails one of the position and shape rules. There are four conditions, and one is enough to leave the egg without a class.

• it touches the image border, so its outline is cut off;
• its area is outside 0.35 to 2.5 times that of a typical egg;
• its solidity is below 0.90, a concave outline that usually means two overlapping eggs;
• it does not look like an ellipse, or its axis ratio is above 1.8.

Solidity measures how far an outline departs from a convex shape. A single whole egg is close to convex, so a low value is a sign of overlap. Touching eggs are separated first by watershed, but no separation is perfect.

Why doesn't the system guess in these cases?

Guessing would have two bad outcomes. An egg cut by the border may have the wrong area and shape, so the dirt measured on it is not reliable. An overlapping pair may be measured as if it were one egg.

Marking the egg as unverified keeps the system honest about what it knows. The egg shows up as an open item, not as a clean egg nobody checked. That also protects audits, because each automatic decision rests on measurements that can be recomputed.

How does manual review fit in?

An unverified egg leaves the automatic decision and goes to a person, who grades it using your line's criteria. Review does not replace calibration: it covers the cases the image cannot settle.

Each per-egg record stores the marking and a plain-language reason, so an audit can see why that egg got no class. The format is described in per-egg records for audits.

If the share of unverified eggs is high, the reason recorded for each egg shows which rule fired. That is the place to look first, whether the camera, the lighting or how the eggs sit on the belt.

How does calibration affect unverified eggs?

The thresholds that decide the marking are fixed configuration values. Calibration with labelled eggs is how you check whether those values fit your eggs, your camera and your light. The method is in calibrating with labelled eggs.

Tracking the unverified rate from the first day gives a baseline. A sudden rise after a change in lighting, camera position or egg colour is a signal to recalibrate, not to adjust the rules on the fly.

Frequently asked questions

Is an unverified egg a bad egg?

Not necessarily. It means the image did not support a reliable judgement, for example because the egg is cut by the border or touches another egg. A person needs to review it before it gets a class, and it may well be a perfectly good egg.

Why doesn't the system grade it anyway?

Because a class based on an incomplete or overlapping outline would be more likely to be wrong. Marking the egg as unverified makes clear what the system knows and what it does not, and the team can see which eggs need a human look.

How many unverified eggs are normal?

There is no universal number. The share depends on how densely eggs sit on the belt, how they are spaced and on the camera. Track the unverified rate on your line from the start, and investigate when it changes by checking the reason recorded for each egg.

Sources

1. Como o Vigiovo classifica ovos sem IA / How Vigiovo grades eggs without AI, Vigiovo