What is reproducibility in egg inspection?

It is the property of the same image, with the same configuration, always producing the same class and the same measurements. In a dispute or an audit, this makes it possible to repeat the decision for each egg and check the result, without relying on the memory of whoever ran the line.

Is a reproducible system a correct system?

Not necessarily. Reproducibility guarantees that the result does not vary, but it does not guarantee that the limit is right for your line. That is why calibration with labelled eggs from your own production is still needed before you trust the classes.

Why doesn't Vigiovo use neural networks?

The choice is one of method. Neural networks change behaviour when they are retrained, while Vigiovo uses classical operations with fixed limits, so that each decision can be explained and repeated. Anyone evaluating other approaches should compare them on their own line.

Explainer · Inspection technology

Why must egg inspection give the same result for the same image every time?

Reproducibility means that the same image, with the same configuration, always produces the same class and the same measurements. In Vigiovo the limits are fixed, no step learns from data and ties follow a defined rule. So each decision can be repeated and checked later, in a dispute or in an audit.

Updated October 9, 2026 · 4 min read

What does reproducibility mean in egg inspection?

Reproducibility is the property of a system giving the same result for the same input. In egg inspection, the input is the image of the egg, and the output is the class together with the measurements that justify it. If the image and the configuration do not change, the result does not change either.

This is different from being right on average. A system can be right most of the time and still not let anyone repeat a specific decision. Reproducibility is about that second property: what was decided can be checked.

Why does this matter for audits and disputes?

When a batch is questioned by a buyer, an auditor or the team itself, the question is: why was this egg graded this way? A reproducible system answers with the same image, the same configuration and the same rule, and reaches the same result.

In human inspection, the decision depends on the operator at that moment, which is hard to repeat. With fixed rules, checking becomes a calculation, not an opinion. The record of each egg that supports that check is covered in the article on per-egg records.

What makes the result reproducible in Vigiovo?

Reproducibility comes from design choices, not from luck. These are the main ones:

The limits are fixed and readable, and no step learns from data.
Ties go to the more severe defect, and the tracking of each egg between frames also follows a deterministic rule.
The scores shown follow a fixed formula, and the highest score always matches the threshold rule.
The record for each egg keeps the class, the scores, the reason and the raw measurements.

Together, these rules make the decision depend only on the image and the configuration. To understand each step, see the how Vigiovo works page.

Are the scores shown probabilities?

No, not in the statistical sense. The scores look like probabilities, but they are a fixed formula applied to the ratio of each defect measurement to its limit. They show how close an egg came to each limit, and there is no trained model behind them.

The class comes from the threshold rule, and the highest score always matches it. An auditor does not need to trust a black box: they can recalculate the same arithmetic from the same measurements.

Does reproducibility guarantee that the result is correct?

No. Reproducibility guarantees consistency, not correctness. If the dirt limit is set for a lighting you no longer use, the system will repeat the same error on every egg. That is why calibration with labelled eggs on your own line comes before you rely on the classes.

There is also a limit to what the camera sees. A reproducible result about the surface is not a result about the inside of the egg. Hairline cracks, the air cell and internal blood spots still require candling.

How do you keep the information needed to repeat a decision?

With reproducibility, checking stays possible even months later. These steps make that real in your operation:

Record the version of the configuration used for each batch, with its start date.
Keep the video or reference images of the batch, together with the per-egg records.
Note any change to lighting, conveyor or limits before it goes into operation.
Repeat the decision for a sample of eggs from each batch with the same configuration and compare the result with the record.

Frequently asked questions

What is reproducibility in egg inspection?: It is the property of the same image, with the same configuration, always producing the same class and the same measurements. In a dispute or an audit, this makes it possible to repeat the decision for each egg and check the result, without relying on the memory of whoever ran the line.
Is a reproducible system a correct system?: Not necessarily. Reproducibility guarantees that the result does not vary, but it does not guarantee that the limit is right for your line. That is why calibration with labelled eggs from your own production is still needed before you trust the classes.
Why doesn't Vigiovo use neural networks?: The choice is one of method. Neural networks change behaviour when they are retrained, while Vigiovo uses classical operations with fixed limits, so that each decision can be explained and repeated. Anyone evaluating other approaches should compare them on their own line.

Sources

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

How Vigiovo helps

Vigiovo grades eggs on the conveyor as clean, dirty, cracked or unverified, in real time, with deterministic computer vision: every decision comes with the measurements behind it, on an ordinary CPU.

See how it works