I am happy to invite you to an online workshop we are organizing on different aspects of large-scale data collection for image quality assessment. To make the most out of the workshop we aim to hold a few different talks over the coming days.
The topic of the talk we are holding on Wednesday the 14th of December at 15:00 Norwegian time (GMT+1) will be “Methodologies for large-scale crowdsourced visual quality assessment”.
The speaker is Mr. Oliver Wiedemann, from the Multimedia Signal Processing Group, University of Konstanz, Germany.
You can join the talk (on zoom) using the following link: https://s.ntnu.no/IQ_workshop
I would appreciate if you can share the event with interested parties you may know.
Visual quality is commonly studied as a concensus-based property by means of subjectively annotated datasets. The gold standard to acquire such ground-truth used to be lab studies, which offer fine-grained control over all relevant influence factors.
In this talk we present crowdsourcing as an alternative approach that scales well even to millions of annotations. We’ll discuss the methodologies that were used in the creation of the Konstanz Visual Quality Datasets and how they affect the resulting annotations.