Search
Menu
Sheetak -  Cooling at your Fingertip 5/24 LB

Heat Maps and Optical Character Verification Take AI-Based Inspection to the Next Level

Jul 18, 2023
Facebook X LinkedIn Email
TO VIEW THIS WEBINAR:
Login  Register
About This Webinar
One subset of deep learning-based machine vision is unsupervised learning. In simple terms, the deep-learning module is trained on images that are defect free. The standard training method consists of training software on defect-free images and images with defects. In this case, every previously encountered defect must be annotated in an image and the module is trained with these images.

The obvious advantage of unsupervised learning is that fewer images are necessary for the training and no annotation is required. The biggest advantage is that it identifies defects of previously unknown shape, pattern, and size. This means that defects that had not been encountered before can be identified.

Opponents of this method note that it may sometimes be difficult to determine why a product was deemed defective. This can be solved with a precise heat map that shows exactly the point in the image where a defect or anomaly was identified. This method is particularly useful for the inspection of products with irregular surfaces or patterns, which is covered in depth in this presentation.

The second significant step in AI inspection was the development of optical character verification (OCV) based on deep learning. Through optical character reading (OCR), the software can read the design of characters on a surface. OCV determines if there are defective letters in the word or the sentence and warns the operator about that letter. Furthermore, the power of the deep learning-based OCR/OCV lies in the localization. Characters in the sentence or word being detected can be anywhere on the image and the AI will locate it, read it, and let the operator know if there are any defects.

*** This presentation premiered during the 2023 Vision Spectra Conference. For more information on Photonics Media conferences, visit events.photonics.com.

About the presenter

Martin BurianMartin Burian was born in 1989 in Kosice, Slovakia where he attended a high school with an IT specialization. After high school, he went to study at a university where he received a Bachelor of Science degree in business administration. That helped him to land a financial controller position at T-systems where he spent two years. Burian then received his master’s degree, studying international business in China, where he spent almost five years. During his study, he worked as a consultant for a company that helps foreigners in China set up their businesses.

After finishing his degree, he became a sourcing manager in a French trading company where he learned valuable skills in trading to serve his personal clients. After a year and a half, he started to work in an e-commerce company as a customer satisfaction and community manager which helped him to further develop his people skills. When the 2020 pandemic started in China, Burian returned to Europe and joined the start-up PEKAT VISION, where he now works as a sales and channel manager.
OCVoptical character recognitionartificial intelligenceinspectionmachine visionheat mapsVision Spectra
We use cookies to improve user experience and analyze our website traffic as stated in our Privacy Policy. By using this website, you agree to the use of cookies unless you have disabled them.