Deep learning inspection software with six AI engines, from anomaly detection to OCR. 2.47ms per 1024×1024 image (segmentation engine, KIPT test).
Deep learning inspection software that finds defects like an expert with six AI engines
It turns experts' OK/NG judgments into training data and inspects everything from film to PCBs with six AI engines on one platform. Models and inspection criteria are bundled into an inspection package (.izpkg) and applied on the line right away.
2.47ms is the segmentation engine's average per 1024×1024 image in the KIPT (Korea Institute of Testing) deep learning inspection test (KIPT-QT2026-008).
59.5%, 1.4× and about 1/3 are company comparisons.
Industries: film · paper · FPCB · electronic PCB · power equipment · electronic components
Six AI engines: Anomaly detection (IntAnomaly) · segmentation (IntSegmentation) · OK/NG classification (IntClassification) · detection (IntDetection) · change comparison (IntComparison) · character recognition (IntOCR), plus optical control with IntCloud Optic
Learns expert criteria: Labeling, data augmentation, training and validation in one screen.
Inspection package (.izpkg): Bundles the model, inspection chain and thresholds into one file to apply on the line and manage versions.
On-premise: Runs on a Windows PC, so production data stays on site.
2.47ms External test average (Segmentation engine · per 1024×1024 image · 100% repeatability · KIPT deep learning inspection test)
59.5% Shorter processing time (Secondary battery can 200→81ms · company comparison)
1.4× Inspections per minute (MLCC 2,934→4,100 · company comparison)
About 1/3 Implementation cost (Versus existing equipment · company comparison)