TALK AT ITU ON “Interspecies Knowledge Transfer for Facial Key point Detection Using Deep Networks” ON THURSDAY

TALK AT ITU ON “Interspecies Knowledge Transfer for Facial Key point Detection Using Deep Networks” ON THURSDAY
Print Friendly, PDF & Email
January 10, 2017

The Computer Science Department of Information Technology University (ITU) is holding a talk on “Interspecies Knowledge Transfer for Facial Key point Detection Using Deep Networks “Smart Internet Technologies” by Maheen Rashid on Wednesday January 11, 2017 at Lecture Theatre 1, 6th floor Arfa Software Technology Park, Ferozepur Road, Lahore from 1500 hrs to 1600 hrs. Entry for interested persons is open.

Guest Speaker Maheen will focus on the importance of detecting facial key points as a necessary step in the automatic detection of expressions of pain in animals and its great potential to positively impact animal welfare.

Her main emphasis would be on the fact that while deep convolutional neural networks have demonstrated impressive performance on a range of computer vision tasks, including human facial key point detection, training deep networks requires large quantities of training data; collecting sufficiently large and diverse datasets of animal facial key points is an expensive and time-consuming process.

This talk will present the challenges in facial key point detection in animals, and a novel method of transferring knowledge from human to animal datasets using deep convolutional neural networks.

Maheen is a PhD student at University of California Davis, where her area of interest is computer vision and deep learning. She works on automating the detection of pain in horses by using deep convolutional networks, and is advised by Dr. Yong Jae Lee. Previously she worked on developing machine learning algorithms for the patient bedside pill scanning device, MedEye, developed by the Icelandic startup Mint Solutions. Maheen completed her Masters in Robotics at Carnegie Mellon University in Pittsburgh in 2014 where she worked on understanding the geometry, layout and composition of images of indoor scenes through the aid of 3D models under the supervision of Dr. Martial Hebert.