The Eye and Vision Sciences research group are delighted to present their seminar series.
The seminars are open to all but please contact firstname.lastname@example.org to reserve your place or for further information.
Wednesday 14 February: Understanding the Deep Learning Revolution: History, Theory and Applications
Speaker: Massimiliano Patacchiola, Centre for Robotics and Neural Systems, University of Plymouth.
Deep learning is a family of machine learning techniques that has recently attracted worldwide attention for its success in speech recognition, natural language processing, and computer vision. It is a matter of fact that deep neural networks are the protagonists of a technological revolution that is changing our lives.
In this presentation, Massimiliano will drive you through the history of deep learning, starting from the logical neurons invented in the Forties and arriving to the latest generative networks able to produce imaginary worlds. We will see how much an artificial neural network resembles its biological counterpart and why it is considered an 'opaque' model. Massimiliano will also show you the other side of the coin. Discrimination, prejudice and erroneous decision making can all arise from biased data. For this reason, there are important ethical questions to answer before deploying deep models in critical applications such as autonomous driving and justice. Finally, Massimiliano will present the results achieved by the CRNS research group in problems such as emotion recognition, pose estimation and drone control, thanks to the use of deep learning techniques.
Wednesday 28 February*: Identifying Impairments of Face Perception with a Novel Clinical Test
Speaker: Andrew Logan, PhD, School of Optometry and Vision Science, University of Bradford.
Faces are amongst the most complex stimuli that the visual system processes. To quantify face discrimination sensitivity, we use synthetic faces, which combine simplicity with sufficient realism to permit individual identification.
We have developed a new clinical test of face perception which is fast (3-4 minutes), repeatable (test-re-test r2=0.795) and can capture normal variability. The Caledonian face test uses an adaptive procedure to measure face discrimination thresholds; the minimum difference required between individual identities for reliable discrimination. A case report of a patient with suspected developmental prosopagnosia indicated that the test is highly sensitive to impairments of face perception (Z-score of -7; c.f. Z-score of -2 for existing face tests).
An investigation of the effect of healthy ageing on face discrimination ability revealed that sensitivity to full faces continuously declined by approximately 13% per decade, after 50 years of age. While older adults performed poorer in every aspect of face perception, there was no effect of age for shape discrimination in an otherwise identical test protocol. This suggests that face discrimination may be particularly vulnerable to the effects of healthy ageing.
Current work aims to quantify the effect of age-related macular degeneration (AMD)- a leading cause of visual impairment in the UK- on face discrimination ability. On average, AMD reduces sensitivity to full faces by a factor of approximately 1.75X. Our data suggest that AMD does not impair discrimination of all face features equally, but disproportionately reduces sensitivity to those which facilitate aspects of non-verbal communication (e.g. facial expressions).
* Refreshments will be served from 15:30 for this seminar.