Coffee & Viz – Deep Neural Networks: Exploration and Practical Applications

This event already happened.
See upcoming events

When

Friday, November 30, 2018
9:30am to 10:30am

Add to Calendar 11/30/2018 09:30 11/30/2018 10:30 Coffee & Viz – Deep Neural Networks: Exploration and Practical Applications

In this talk Dr. Healey, Goodnight Distinguished Professor in NC State's Department of Computer Science, will discuss the area of deep learning, with a specific  focus on convolutional neural networks (CNNs), a class of deep neural  network often used for image analysis. The talk will begin with a very brief  overview of the deep learning area, followed by a more focused  discussion of CNNs. Dr. Healey will then present two ongoing projects in his  research laboratory: a method to visualize the internal structure and  response of a CNN to specific inputs or classes of inputs, and a system that can generate content for faces that have significant areas "removed" by a grey mask. The face completion can not only generate a unique, photorealistic repair; it can also control properties of the result, for example, whether the individual is smiling of frowning, whether (given sufficient ambiguity in the original, masked image) the individual looks like a male or a female, and so on. The talk will conclude with examples, both positive and negative, of the face completion system.  

 

Teaching and Visualization Lab

Where

Event Description

In this talk Dr. Healey, Goodnight Distinguished Professor in NC State's Department of Computer Science, will discuss the area of deep learning, with a specific  focus on convolutional neural networks (CNNs), a class of deep neural  network often used for image analysis. The talk will begin with a very brief  overview of the deep learning area, followed by a more focused  discussion of CNNs. Dr. Healey will then present two ongoing projects in his  research laboratory: a method to visualize the internal structure and  response of a CNN to specific inputs or classes of inputs, and a system that can generate content for faces that have significant areas "removed" by a grey mask. The face completion can not only generate a unique, photorealistic repair; it can also control properties of the result, for example, whether the individual is smiling of frowning, whether (given sufficient ambiguity in the original, masked image) the individual looks like a male or a female, and so on. The talk will conclude with examples, both positive and negative, of the face completion system.  

 

Contact Information

Shelby Hallman
(919) 513-3661

Admission Information

Free and open to the public.

Accessibility

If you require a disability-related accommodation to participate in this event, please contact Marian Fragola at marian_fragola@ncsu.edu or 919-513-3481 to arrange services. Early requests are strongly encouraged to allow sufficient time to meet your access needs​.

flier for event