Telepresence Video Quality Assessment

Video conferencing, which includes both video and audio content, has contributed to dramatic increases in Internet traffic, as the COVID-19 pandemic forced millions of people work learn from home. Global traffic conferencing dramatically increased Because this, efficient accurate quality tools are needed monitor perceptually optimize telepresence streamed via Zoom, Webex, Meet, etc.. However, existing models limited their prediction capabilities on multi-modal, live streaming content. Here we address significant challenges Telepresence Quality Assessment (TVQA) several ways. First, mitigated dearth subjectively labeled data by collecting $$\sim $$ 2k videos different countries, crowdsourced 80k subjective labels. Using this new resource, created a first-of-a-kind online framework for streaming, using multi-modal learning with separate pathways compute visual predictions. Our all-in-one model is able provide predictions at patch, frame, clip, audiovisual levels. achieves state-of-the-art performance databases our TVQA database, considerably lower computational expense, making it an attractive solution mobile embedded systems.