AI knows when you are happy - Understanding the process of facial recognition through AI

AAII Technical Lecture Series (ATLS-4)

Human emotions and their modeling are crucial aspects in developing intelligent systems. With the growth in multimedia applications, including social robots, medical treatment, and video chat, facial emotion recognition (FER) has become an important but hard problem to solve. In this introductory talk, we will learn how to recognize human emotions from images and videos using computer vision and machine learning (in particular deep learning).

In the fourth session of ATLS, Dr. Debanga Raj Neog, the co-founder and CTO at Nytilus, Canada, a computational imaging startup for industrial applications, was our respected speaker. Focusing on understanding the process of facial recognition through AI, Dr. Neog talked us through this vast discipline with an easy to grasp presentation with explanations to each slide presented.

He started the session by sharing his journey from IITG to founding an AI startup in Canada. While pursuing his bachelor’s degree, Dr. Neog got the chance to visit various countries as part of his summer internships and discovered his keen interest in higher studies and research.

Our technical talk started with Dr. Neog giving us a background on why it has become crucial nowadays for AI to understand and judge human emotions. Human emotions and their modeling are critical areas in developing intelligent systems. With the growth in multimedia applications, including social robots, medical treatment, and video chat, facial emotion recognition has become an important but hard problem to solve. He then introduced us to the various human emotions models and theories proposed by various scientists over the years. This was followed by him briefing us on the various methods of measuring human emotion, human emotion in images, classification of facial unit estimation, facial emotion estimation, and end-to-end DotNetNuke. He concluded his talk by giving us a few examples of simple to advanced networks in deep learning, classifying it in Vanilla Transfer Network with examples of ResNet, EfficientNet, and explaining the use of Google FEC and RingNet as tools for facial expression recognition and identification.

Related links:

• Youtube Link of the talk: ATLS 4
• Presentation slides
• Books:
  • Computer Vision: Models, Learning, and Inference
  • EfficientNet
• Papers
  • ResNet
  • Google FEC
  • RingNet
• Courses
  • Machine learning: Coursera

  • Computer vision: Udacity

    CS231n.standford.edu

• Start learning hands-on!
  • Computer vision library: OpenCV
  • For ML practice: Kaggle

Speaker:

Dr. Debanga Neog did his B.Tech in Electronics and Communication Engineering from the Indian Institute of Technology Guwahati (IITG) and Ph.D. from The University of British Columbia (UBC), Canada. He was the recipient of the Mitacs-INRIA Globalink Research Award and Four-Year Doctoral Fellowship at UBC. He has 12+ years of experience in leading academic and industrial projects in the areas of computer vision, computer graphics, and machine learning. Contact:

• LinkedIn
• Personal Site
• Google Scholar

Speaker: Dr. Debanga Raj Neog, Co-founder & CTO, Nytilus, Canada (BTech, IIT Guwahati | Ph.D. from University of British Columbia)
Topic: AI knows when you are happy 
Date: 31 January 2021 
Time: 6 PM (IST)