Industries›Manufacturing

Vision AI for
Manufacturing

Various deep learning technologies are being utilized to enhance the extent and versatility of automation within the inspection process.

Specifically, deep learning-based models are employed to complement the limited flexibility and scalability of rule-based and machine learning models.

Inquiry

Manufacturing

Use Cases

PCB Classification model to minimize false positive* (resulted by Machine Learning Model)

Issues and
Needs to Address

  • Minimizing false positive issues of ML-based models that classify normal products in the existing inspection process as defective.
  • Ensuring the stability of performance in new deep learning-based models through training techniques is necessary.

Solutions and
Applied Technologies

  • Using SNUAILAB's AutoCare, we re-trained the inspection model (SINNet, SNUAILAB Inspection Network) with on-site data.
  • To compare and contrast the performance of the retrained SINNet with the existing ML model, we applied AutoCare Edge and conducted A/B tests, visualizing the performance history of the retrained SINNet.

Results of Technology Implementation

  • 01The incidence of false positive* issues in existing ML models has been reduced by approximately 50% compared to the previous iteration.
  • 02AutoCare supports the stable maintenance of performance for the newly implemented SINNet-based DL model, adapting to changes in the field.

* False Positive refers to a group of genuine products erroneously classified as defective by existing models.

Are you curious about SNUAILAB's AI technology
applied across various industries?

Maximize your business values
by adopting industry-specific AI solutions.

Inquiry