LUXIO - AI· VISUAL QUALITY INSPECTION · PART OF THE PANDORA PLATFORM
Vision AI that inspects EVERY UNIT unit on your line — at production speed.
LUXIO - AI
Luxio-AI is HEXA IoT's machine-vision quality inspection system. It watches your production line through a camera, judges each product against a model trained on your own goods, and acts on the result — rejecting defects, flagging contamination, and recording every decision in Pandora.
Quality control that doesn't get tired, distracted, or inconsistent
ai Solution for Real-Time Monitoring
Human visual inspection is slow, fatiguing and varies from person to person and hour to hour. On a fast line, that's where defects and contaminants slip through to the customer.
Luxio-AI replaces that with a camera and a trained deep-learning model that inspects 100% of production in real time — not a sample. It captures an image of each unit, classifies it against the patterns it has learned from your good and defective products, and triggers an action: pass, reject, or alert. Because it's built on object-detection models rather than fixed rules, it recognises products regardless of size, shape or position, and holds up under the variation of a real factory.
It is fully integrated with the Pandora platform, so every inspection isn't just a pass/fail on the line — it's a logged, traceable record you can audit later.
Capabilities
-It is IoT remote monitoring & control via Cloud Technology
-It has auto-triggering system by pre-set parameters
-It able to train to sort different types of products
-It is fully customizable rejection system to suit different products needs
Quality control that doesn't get tired, distracted, or inconsistent
IoT Solution for Real-Time Monitoring
Human visual inspection is slow, fatiguing and varies from person to person and hour to hour. On a fast line, that's where defects and contaminants slip through to the customer.
Luxio-AI replaces that with a camera and a trained deep-learning model that inspects 100% of production in real time — not a sample. It captures an image of each unit, classifies it against the patterns it has learned from your good and defective products, and triggers an action: pass, reject, or alert. Because it's built on object-detection models rather than fixed rules, it recognises products regardless of size, shape or position, and holds up under the variation of a real factory.
The system uses IoT remote monitoring and control via cloud technology, and it can be integrated into other systems with various types of hardware.
It is fully integrated with the Pandora platform, so every inspection isn't just a pass/fail on the line — it's a logged, traceable record you can audit later.
Capabilities
-It is IoT remote monitoring & control via Cloud Technology
-It has auto-triggering system by pre-set parameters
-It able to train to sort different types of products
-It is fully customizable rejection system to suit different products needs
LUXIO captures visual images via camera in a preset condition, identifies them according to training modules, and assign numeric properties to it. User is able to program actions according to the visual analytics.
From product to decision in milliseconds
Every Luxio-AI installation runs the same five-stage pipeline, engineered end-to-end for the specific inspection task — from lighting and optics through to the reject mechanism and the record in Pandora.
Every Luxio-AI installation runs the same five-stage pipeline, engineered end-to-end for the specific inspection task — from lighting and optics through to the reject mechanism and the record in Pandora.
Trigger & capture
A sensor detects the unit in position and triggers the camera to capture a clean, consistently-lit image.
Pre-process
The image is normalised for lighting and alignment so the model sees consistent input every time.
Inference
The trained model detects and classifies the product — good, defective, contaminated or mislabelled.
Act
Pandora triggers the outcome — an air-jet or arm rejects the unit, or an alert is raised — via PLC.
Log
The result, image and timestamp are recorded in Pandora for OEE, traceability and audits.
WHAT IT INSPECTS
Three jobs that protect your product and your reputation
Because the object-detection algorithm is fully customisable, Luxio-AI is trained per product line — so the same system can sort very different goods, each against its own quality standard. The rejection logic is equally configurable: set the parameters that define a fail, and choose what happens when one is found.
Defects
Catches off-spec and damaged units before they're packed and shipped.
-Cracks, scratches, deformation
-Wrong size, shape or colour
-Malformed or incomplete product
Contamination
Flags foreign material and anomalies that must never reach a customer.
-Foreign objects on the line
-Visual anomalies vs a known-good product
-Critical for food safety & recalls
Labelling & print
Verifies the right label is present, placed and legible.
-Missing, wrong or skewed labels
-Print & date-code presence
-Evidence for compliance audits
How HEXA IoT implements it
From your factory floor to a running system
Luxio-AI isn't an off-the-shelf box you're left to configure. As an operator-proven provider, HEXA IoT implements it as a managed deployment — and the model is trained on your products, because no two lines look the same.
Assess the line
We study the product, the line speed, the defects that matter, and the physical constraints — then design the camera position, lighting and reject mechanism for that specific task.
Gather training images
We collect images of good product and the defects, contaminants and labelling errors you care about. The quality and range of this dataset is what determines real-world accuracy.
Train & validate the model
We train a custom object-detection model on your dataset and validate it against held-out images, tuning until it reliably separates good from bad under your line's real conditions.
Integrate with the line & Pandora
We connect the camera, edge unit and reject mechanism to your PLC, and wire the results into the Pandora platform for live monitoring, OEE and traceability.
Run, monitor & improve
As part of the managed service, we monitor performance and retrain the model as products, packaging or defects change — so accuracy doesn't drift over time.
Part of the Pandora platform
Every inspection becomes a traceable record
On its own, a vision system makes a pass/fail decision and moves on. Integrated with Pandora, Luxio-AI turns each decision into data — feeding the same dashboard that runs the rest of your floor, and connecting upward to your ERP, MES and SCADA.
CAPTURE
Camera + edge inference at the line
low latency · works offline
CAPTURE
PLC triggers reject / alert
configurable rejection logic
CAPTURE
Pandora logs result, image & time
pass rate · defect types · OEE
CAPTURE
Linked to batch / lot for traceability
audit & recall ready
CAPTURE
Up to ERP / MES / SCADA
via REST API · OPC UA
Why it matters
What automated vision QC changes on the floor
100% inspection
Every unit checked, not a sample — closing the gap where costly escapes happen.
Consistent standard
The same judgement applied to the first unit of the day and the last — no fatigue, no drift.
Evidence on record
A logged, traceable QC history ready when a customer or auditor asks.
AT A GLANCE
Luxio-AI capabilities
CAPABILITY
DETAIL
Inspection type
Deep-learning object detection & classification — defect, contamination, labelling
Coverage
100% of production in real time (not sampling)
Training
Custom model per product line, trained on your good & defective images
Adaptability
Recognises products regardless of size, shape or position
Triggering
Sensor-triggered capture, synchronised to line motion
Compute
Edge inference at the line for low latency; cloud sync for monitoring
Rejection
Fully configurable parameters & reject mechanism (air-jet, arm, diverter) via PLC
Integration
Pandora platform; ERP / MES / SCADA via REST API & OPC UA
Records
Result, image & timestamp logged; linked to batch/lot for traceability & OEE
Delivery
Managed service — assessed, trained, integrated, monitored & retrained by HEXA IoT
Funding
Qualifies under SIRIM Smart Tech Up + 200% automation tax deduction
COMMON QUESTIONS
About implementing vision AI on your line
Do I need to know what defects to look for?
No — part of our assessment is identifying the defects, contaminants and labelling errors that matter on your line, then building the training set around them.
Will it work with my existing line and machines?
Yes. Luxio-AI is added to your line with its own camera, lighting and edge unit, and connects to your existing PLC and reject mechanisms — no need to replace equipment.
What happens when my product or packaging changes?
As part of the managed service, we retrain the model so it keeps recognising your current product accurately, rather than degrading over time.
Does it keep working if the internet goes down?
Yes — inference runs on an edge unit at the line, so inspection and rejection continue offline. Results sync to the cloud once connectivity returns.
How is this different from a normal camera or rule-based vision?
Rule-based systems only catch what they're explicitly programmed to see and break with small variations. Luxio-AI learns from examples, so it handles the natural variation of a real production line and can be trained for new products.