MLL Distinguished Lecture Series: Al Bovik
3:00 PM – 4:00 PM CT
3:00 PM – 4:00 PM CT
The Machine Learning Lab will be hosting the 2020 Technology and Engineering Emmy(r) Award winner, Al Bovik, as part of the MLL Distinguished Lecture Series:
Why is Blind Video Quality Prediction So Hard?
Speaker: Al Bovik
Professor, Cockrell Family Endowed Regents Chair in Engineering at The University of Texas at Austin
Friday, March 05, 2021
3:00 PM - 4:00 PM CT
Virtual: Register here: https://utexas.qualtrics.com/jfe/form/SV_74EJWfCd3jD38Me
Every year, billions of videos are captured by inexpert users and shared on social media, but are distorted, reducing their desired perceptual quality. Numerous distortions can arise: blurs, compression, jitter, shake, noise, judder, over/under-exposure, etc., often combining to create multitudes of composite impairments impossible to model analytically. Importantly, perceived quality is not only a function of amount of distortion, but also of interactions between distortion and content. For example, videos having identical applied distortions might lie at the opposite ends of the perceptual quality scale, because of neurophysiological masking processes.
We will explore why video signals are "special," with internal statistical structures that visual systems have optimally evolved to optimally encode and process. These special attributes have been used to create video quality models that now monitor and control much of all Internet traffic. We'll talk about the need for very large-scale psychometric databases of human judgments of video quality, including our latest, and our most recent deep video quality predictor, which uses parallel 2D and 3D ResNets to feed a temporal InceptionTime inferencer. The results suggest that the blind video quality prediction problem, long regarded as impossible, might indeed be practically solvable.
Speaker Bio: Al Bovik is the Cockrell Family Regents Endowed Chair Professor at The University of Texas at Austin. His research interests land squarely at the nexus of visual neuroscience, deep learning, and digital streaming and social media. His many international honors include a 2020 Technology and Engineering Emmy(r) Award, the 2019 Progress Medal of the Royal Photographic Society, the 2019 IEEE Fourier Award, the 2017 OSA Edwin H. Land Medal, a 2015 Primetime Emmy(r) Award from the Academy of Television Arts and Sciences, and the Norbert Wiener and 'Sustained Impact' Awards of the IEEE Signal Processing Society.
Contact us at ML-Lab@Austin.utexas.edu