Resources
Resources
The questions that actually come up when evaluating inspection, answered the way our engineers answer them on site.
- Getting started
How many defect samples do you actually need?
How many defect images do we have to collect before AI inspection can start?
Five is the norm. Where a product has no defect history, or is too large to bring to us, a model can be built from good samples alone — or from a single image. The assumption that you need thousands belongs to the rule-based era.
Read in full → - Technology comparison
Rule-based AOI versus deep-learning inspection
We already have AOI equipment. Is there a reason to move to AI inspection?
The difference shows up when the lighting shifts or a new defect appears. Rule-based means rewriting the algorithm; deep learning means training it. The case is strongest on lines where over-detection already sends work back to a human.
Read in full → - Security
Where does inspection data live?
Does our line imagery leave the building — and does it leave the country?
It does not. DSTEK deploys on-premise by default: training, inference and storage all run on customer infrastructure. Cloud connectivity is configured only when the customer asks for it.
Read in full →
We do not add entries to fill a page. Topics are published as they clear internal review.