| Standard | Qualcomm QCS5430 · 3.5 TOPS |
| Advantech iCAM300 | Qualcomm QCS6490 · 12 TOPS |
| Application | Object detection & defect identification |
| Processor | Qualcomm QCS5430 · AI 3.5 TOPS |
| Memory | LPDDR4X 4GB + UFS 2.x 64GB |
| Camera Module | 8 Mega Pixels |
| Supported OS | Linux |
| Audio | 1 digital microphone / beeper |
| I/O Interfaces | Ethernet ×1 · Power key · Link LED ×1 · Power LED ×1 · Indicator LED ×1 · Expansion MicroSD · HDMI · Power plug DC12V / 2.5A |
| Sensor | 1/1.8″ IMX678-AAQR1-C |
| Active array size | 8M / 3840H × 2160V |
| Connect Type | MIPI |
| Interface Type | FF |
| View angle | 135° |
| Depth of Field | 0.79 m ~ inf |
| Module Dimension | 20 × 20 × 44.44 mm |
| Pin Output | 34 PIN |
| Module Slave Address | 0x34(W) / 0x35(R) |
Quality assurance and QA departments in manufacturing · lines with a manual inspection bottleneck · companies that want AI but have no AI team · plants that need to raise yield and efficiency quickly.
Cosmetic inspection (scratches, contamination, defects) · assembly verification (missing or wrong parts) · intelligent classification (product sorting) · operator poka-yoke (preventing incorrect operation).
Rapidly identify surface defects and dimensional anomalies on machined metal parts — scratches, cracks, deformation — and output a control signal in real time to reject or alarm.
Recognise in real time whether parts are correctly fitted, and use I/O to alarm or stop the equipment directly — the first choice for poka-yoke on assembly lines.
Accurately identify object type and position and sort by rule in real time, reducing error rates — suited to mixed-model lines.
Monitor the process and operator actions in real time, check that work follows the standard procedure, and catch missed or incorrect steps on the spot.
Automatically identify the type and condition of recyclable items and control the equipment to sort and divert them — high-speed sorting with less reliance on labour.
Identify position and object from the image and control the robot arm to pick or operate directly, improving positioning accuracy and lowering the development barrier.
Clarify the inspection target, the takt time and the conditions on site.
Data collection, labelling and training in 1–3 days.
Mounting, I/O wiring and testing in one day.
Full production, iterated onward by your own people.
This product is part of the government's AI adoption framework and can be combined with government subsidies and enterprise adoption programmes to speed up AI deployment.
An AI adoption tool, listed on the IDA's AI Tools Cloud Marketplace.
LISTEDCertified artificial intelligence technology service provider.
CERTIFIEDAdvisory support and solution recommendation.
ENDORSEDThe edge AI vision market is taking off. We are looking for partners around the world who understand the factory floor, to bring the AI Action Camera to their local production lines — with full technical, training and market support from us.
No-code, live in 10 minutes, and no change to the customer's existing PLC — a product experience where the demo closes the deal, backed by Taiwanese government certification and a listing on the official AI Tools Cloud Marketplace.
Distributor training, a sales toolkit and demo units, with FDE engineers helping you land the critical early deployments.
Automation system integrators (SI), machine vision and industrial camera distributors, industrial equipment resellers — with local manufacturing channels and technical service capability.
※ Static prototype — the WordPress version is built with Fluent Forms and delivers to sale@prefactortech.com
The partners bringing Physical AI to factory floors around the world with us.
Bring us your inspection problem — 1–3 days to build the model, results within a week.