AI & Deep Learning for Vision Applications

July 10, 2018

12:00-1:30 pm ET

ABOUT THIS WEBINAR

ML for verification and defect detection/classification
Image based ML algorithms have been advancing exponentially, but manufacturing industries have been slow to adopt these new methods.  Supervised ML algorithms can be taught to perform inspections with less developer effort for certain classes of problems when compared to traditional methods, and will often work where traditional techniques are infeasible.  Advances in IIOT have reduced the cost of acquiring and labelling large datasets, situating manufacturing in a sweet-spot for leveraging ML.

Localization for debugging and more
A major criticism of ML is its “black-box” nature.  In convolutional neural network (CNN) based ML models one very useful technique is the use of “class activation maps” – a heatmap showing the area(s) of an image that most heavily affected the output of the model for a certain image.  By inspecting high-activation areas we can understand why a model made a certain prediction the way it did.  Looking further, these algorithms can be used to localize parts without explicit training.

Continuous Improvement
Vision inspection systems in manufacturing generate a lot of data. In many cases any incorrect predictions made by the system can be identified and associated with recorded data, allowing continuous-training and algorithm improvements to be had almost for free.  In contrast with traditional vision systems which often degrade over time due to changing environmental and physical conditions, an ML based vision system can continually improve the longer it runs.
 

Webinar Image

SPONSORS:

WEBINAR PRESENTERS
Image of Nick Stupich

Nick Stupich

Nick Stupich works at Prolucid, where he helps clients with their computer vision and machine learning software solutions. His work has spanned many industries, including medical, automotive, energy and manufacturing. Prior to working at Prolucid he developed software to analyze and recognize gestures from electromyography signals, and correct rolling shutter distortion in videos. Nick has a Bachelor’s degree in Engineering Physics and a Master’s degree in Electrical and Computer Engineering from Carleton University. When he’s away from his computer, Nick likes to ski, bike, and run up and down mountains.

Search AIA:


Browse by Products:


Browse by Company Type: