
The PEKAT VISION Anomaly Detector module is able to detect previously unseen or novel defects, greatly reducing training time and streamlining the deployment process. This makes it an ideal choice for industries that require quick, reliable, and efficient inspection solutions.
Anomaly Detection Based Solely on Defect-Free Images

Easily Identifies Defects and Anomalies
The Anomaly Detector module identifies deviations from the OK state that indicate defects, useful particularly in applications where defects are rare and varied. Even when the shape, size, location, or type of defect is unpredictable, the Anomaly Detection module can easily detect these defects, ensuring comprehensive quality control even in complex and variable conditions.

Easily Adapts to Natural Variations in Materials
PEKAT VISION’s proprietary Focused-learning Technology offers unparalleled accuracy in distinguishing anomalies while accommodating natural variations in complex patterns.
How the Anomaly Detector Works
The Anomaly Detector, like other PEKAT VISION deep learning modules, uses a multi-layer neural network to automatically learn features from data, excelling in complex and variable inspection tasks. Unlike supervised methods, which rely on labeled datasets to classify images, objects, or defects, the Anomaly Detector is an unsupervised approach trained on defect-free images.
This method identifies deviations from the norm, highlighting potential defects without needing detailed annotations. Users simply assign an ‘OK’ label to at least one defect-free image, and the module then detects anomalies based on this reference model, which it continuously uses to inspect new images for any defects.
Watch the Anomaly Detector Tutorial
Watch a step-by-step video tutorial for the PEKAT VISION deep-learning Anomaly Detector module. From basic setup to troubleshooting and calibration. It covers all you need to know to get started and optimize your defect detection processes.