Skip to main content

Products

We stay true to the present and prepare for the future in a fast-changing IT era.

Solution Overview Introduction
AiEYE ADC

AiEYE ADC is an automatic defect classification solution that uses AI to analyze SEM images and wafer maps, finding and classifying more than 180 defects a minute.

Built from AI technology and a wealth of training data gathered on semiconductor production lines, AiEYE ADC
is a core tool for analyzing the causes of defects - and the key to higher yield and lower cost.

Finding and classifying defects is the precondition for raising yield and cutting cost in semiconductor manufacturing. Working out why a defect occurred and resolving it is arduous work, and automation has so far had its limits.
Drawing on AI video analysis expertise and methodology built up through the trust of our customers, SOFTonNET developed AiEYE ADC to be low-latency, high-performance and fast.

  • SEM Images and Wafer Maps
    Supported Together

    Built on SOFTonNET's AiEYE platform, AiEYE ADC brings together the Layered Defect Classifier, which analyzes SEM images, and the Wafer Map Classifier, which classifies defects on wafer maps.

  • Accurate, Fast
    AiEYE Analysis

    AiEYE ADC classifies wafer map defects with 98% accuracy. The deep learning analysis model can find and classify 180 defects a minute.

  • An Integrated Training System
    for Continual Improvement

    Improving a process and raising yield call for defect detection that keeps getting better.
    AiEYE ADC includes not only the detection and classification system but a deep learning training system as well, so the analysis model can go on improving how it finds and classifies defects.

Wafer Map Classifier
The Wafer Map Classifier in AiEYE ADC uses training data augmented with a GAN (Generative Adversarial Network) to classify wafer map defects with up to 98% accuracy.
Wafer map defect classification by the Wafer Map Classifier
Layered Defect Classifier
The Layered Defect Classifier in AiEYE ADC analyzes SEM images and finds common defects with 95-100% accuracy. Defects that are hard to recognize - SiGe growth process defects, fine patterns - are classified with better than 90% accuracy.
SEM defect images classified by the Layered Defect Classifier
Features Features

AiEYE ADC brings together SOFTonNET's AI video analysis expertise and the methodology to apply it,
offering the chance to automate wafer defect detection and classification at lower cost.

  • Centralized Reading

    AiEYE ADC reads centrally, with uninterrupted service and load distribution in mind.
    Defect detection and classification run on one or more GPU-equipped servers to give users the best possible performance.

  • A Web-Based User Interface

    Users review classified defects on a web-based screen.
    The user management policy of the AiEYE platform can restrict the view to the readings for the lines assigned to that user.

  • Easier Integration
    with Process Solutions

    AiEYE separates analysis - detection and classification - from the user interface, and that structure makes it easier to combine with the process solutions you already run.

  • Model Deployment Without Downtime

    Defect detection has to continue for as long as the production line is running.
    Docker-based management lets AiEYE ADC deploy and update the analysis models produced by the training system without interrupting the service, so a new model reaches the line immediately.

How the AiEYE ADC solution works
How the AiEYE ADC solution works