Welcome to the 27th PyData Cambridge virtual meetup!

This is a virtual meetup. We will use zoom, login details published closer to the date.

Agenda

19:00 - Introduction
19:15 - "Active learning, scaling and dataset merging for ranking and rating surveys in image quality assessment" -- Aliaksei Mikhailiuk
19:45 - End

Code of Conduct

PyData is dedicated to providing a harassment-free event experience for everyone, regardless of gender, sexual orientation, gender identity, and expression, disability, physical appearance, body size, race, or religion. We do not tolerate harassment of participants in any form.

The PyData Code of Conduct governs this meetup. ( http://pydata.org/code-of-conduct.html ) To discuss any issues or concerns relating to the code of conduct or the behavior of anyone at a PyData meetup, please contact NumFOCUS Executive Director Leah Silen (leah@numfocus.org) or organizers.

Talk

** Active learning, scaling and dataset merging for ranking and rating surveys in image quality assessment **

Abstract:

Automatic or objective assessment of image and video quality is a stepping stone for developing accurate compression, reconstruction, enhancement and tone-mapping algorithms. Since the ultimate consumer of visual content is a human the results of these algorithms must be perceptually pleasing. The final quality assessment model built for this type of applications should thus show high correlation with subjective quality as perceived by human observers. The major obstacle to developing an “ideal" model is the lack of a sufficiently diverse training dataset. Training data for image and video quality assessment is hard to obtain, as it requires running expensive and time-consuming experiments with human observers. Thus, existing subjective image quality datasets are homogeneous and fragmented. In the talk I will go through the whole development pipeline for image quality assessment algorithms from active learning for data acquisition and dataset merging to modelling human judgements. Even though examples in the talk are based on image quality, methods discussed are widely applicable to cases where scaling of the conditions is necessary, for example consumer surveys, game tournaments, and education.

Bio:

Aliaksei has completed his PhD last year under the supervision of Dr. Rafal Mantiuk at the Computer Laboratory, University of Cambridge. Currently he continues his research focused on applications of Machine Learning to the problem of modelling Human Visual System. Before his PhD he worked on Deep Learning algorithms for multi-dimensional data reconstruction under the supervision of Dr. Anita Faul. Before joining Cambridge he worked on Computer Vision and Machine Learning algorithms for pothole detections at the University of Bristol.