AI defect detection
About
Neureili employs an unsupervised AI defect detection system.
No AI experts are required; on-site personnel can complete high-quality inspection model training in just a few minutes.
Common AOI issue
Traditional AOI relies heavily on rules and labeled data. If the software's accuracy is improved, the overkill rate (Overkill) and false positive rate become extremely high, which increases the extra manpower needed to re-inspect misclassified items.
In supervised learning defect detection AI, defects that have not been labeled and learned by the AI will go undetected.
Other AOI issue
With other AI solutions, a large number of defective product samples need to be collected so that the software can learn to identify defects. If the manufacter’s production yield is high, collecting defective items can be difficult and time-consuming.
Experienced quality inspectors also need to spend a lot of time manually labeling defects on the computer. If the labeling is inaccurate, the AI’s performance will be affected.
Feature
5-10 images
Provide a small number of defective product/normal samples to train the model (the exact number of images depends on the complexity of the case)
5 mins
Using unsupervised learning, you can quickly train AI models and use the models on production in a short time.
100ms
Detection and response time can be as fast as 100 milliseconds
(depending on the complexity of the detection)