TRAIN · AIVQC Trainer

Before a model reaches the line, it goes through Trainer.

Trainer is the process engineer's workstation — where data and models are built and verified before anything is approved for publishing.

Inspection quality starts with data, not with the model. In Trainer you collect and annotate OK and NOK images, build a representative training set and an independent test set the model has never trained on.

Trainer lets you train several model iterations and compare them side by side on the same test data — instead of relying on a single, unverified version. Only once an iteration meets the agreed acceptance criteria can it move on to Server as an approved package.

This component is used by an engineer or integrator, not by a line operator — it runs on an engineering workstation, outside the rhythm of production, so experiments and comparisons don't affect the ongoing inspection.

Images and annotations

Import production images, label defect classes, and build a training set and an independent test set.

Training and comparison

Train successive model iterations and compare their results side by side on the same test data.

Acceptance criteria

Decision thresholds and quality requirements are agreed before training, not fitted to the result afterwards.

Data versioning

Every iteration of the dataset and the model is traceable — you know exactly what the model was trained on.

PLACE IN THE PLATFORM

Trainer prepares what Server later publishes.

A model version approved in Trainer becomes a package that AIVQC Server distributes only to selected stations. Trainer never connects directly to the production line — that separation between experimental work and the production environment is deliberate.

FAQ

AIVQC Trainer

Does Trainer need our own data, or do you help collect it?+

During the pilot we jointly define what images are needed and help plan how to collect them. In ongoing operation, production data is collected by your team under the agreed imaging conditions.

How many model iterations fit in a typical cycle?+

During the pilot we train up to three main iterations. In ongoing process work, the number of iterations depends on how quickly new, difficult cases appear.