Flawless Production Thanks to AI
www.posiv.cz ǀ 27
24 VISION vznikla v roce 2019 a nabízí řešení pro
komplexní kontrolu kvality. Její systém využívá
neuronové sítě a umělou inteligenci. Detekuje vady
a kontroluje konguraci výrobků v reálném čase
s deklarovanou spolehlivos 99 %.
24 VISION was founded in 2019 and provides
soluons for complex quality control. Its system
leverages neural networks and arcial intelligence
to detect defects and verify product conguraons
in real me, with a declared reliability of 99 %.
na výrobní systémy zákazníka pomocí standardizova-
ných integrací, které jsou součástí našeho produktu
jakožto aplikační rozhraní. V okamžiku, kdy je výrobek
na analytickém stanovišti, náš systém obdrží – nebo
si případně vyžádá – potřebné informace o typovosti
výrobku a spustí pouze ty analýzy, které jsou validní.
Variabilita skupin produktů je definována v našem sys-
tému a není nijak omezena.
Unikátnost systému
Systém vidí nejen defekty, ale i konfigurační rozlože-
ní – rozezná správnou montáž i následnou kontrolu
povrchu. Všechny tyto činnosti spolu souvisí a mají
vydefinovaný přesný popis. K detekci využíváme různé
druhy neuronových sítí a kombinujeme je mezi sebou.
Výsledkem je obrovská variabilita detekcí, kterou navíc
průběžně doplňujeme o nové typy neuronových sítí.
Flawless Producon Thanks to AI
ANNIE, the visual inspecon system developed by the Czech start-up 24 VISION, is already
running in full operaon at major automove players — and that’s just the beginning. Their
approach to integrang arcial intelligence and computer vision directly into producon sets them
apart from the compeon. What exactly does it oer, and how does it work, given that it can
be connected to exisng lines and easily congured – without any programming?
Quality Control in Practice – ANNIE
We are continuously working to efficiently
incorporate technological changes into
ANNIE, based on customer feedback,
while at the same time driving innovation
with our own ideas inspired by the market
and technological progress.
Our priority is to ensure that customers can
actively work with the product themselves.
Thanks to user-friendly tools, they can
configure ANNIE independently, expand
it to other parts of production, and make use
of its results — all without any programming
knowledge.
What sets us apart most from
the competition is our approach to deploying
AI and Computer Vision for quality control
directly in production. Our solution can
be used with cameras or sensors from other
manufacturers, as ANNIE does not rely
on specific analytical methods embedded
in the cameras themselves, but simply
processes their output — the image.
Expanding Beyond Automotive
We are beginning to enter the fields
of electronics and mechanical engineering.
There is significant potential, especially
in precision engineering, where not only
dimensions but also subtle visual defects
in materials or machining are assessed.
These defects are extremely difficult
to detect with the human eye, and to boost
production efficiency, customers also need
precise records of defect types and the
ability to ensure full traceability.
In the coming years, we plan to focus more
intensively on the pharmaceutical industry,
which holds great potential for our solution.
In general, our product can be applied
anywhere exceptionally high product
quality is required, particularly in large-scale
manufacturing.
Quality Control for Products With High
Variability and Complexity
Our solution was developed specifically
to handle variability and complexity
efficiently. Unlike competing systems,
we deliberately avoid focusing on simpler
use cases where only a few defects need
to be detected or a single type of surface
flaw is analysed.
Our strength lies in the dynamic processing
of product input data at the moment
of detection and in connecting to the
customer’s production systems via
standardised integrations, which form part
of our product as an application interface.
When a product reaches the inspection
station, our system receives — or, if necessary,
requests — the required information about
the product type and launches only those
analyses that are relevant. The variability
of product groups is defined within our
system and is not limited in any way.
System Uniqueness
The system recognises not only defects
but also configuration layouts — it can verify
correct assembly and carry out subsequent
surface inspections. All of these tasks are
interconnected and defined with precise
descriptions. For detection, we use various
types of neural networks and combine
them with one another. The result is a vast
range of detection capabilities, which
we continuously expand with new types
of neural networks.
StartUp