CMSD and the Temple of Doom: Escaping VQA's Legacy Traps
summary
Implementation uses CMSD-MQA to push encoder-calculated per-segment scores via CDN headers so the pipeline sees VMAF SSIM and PSNR at the source rather than relying on downstream probes. The claim is concrete: mean time to detection down from 20+ minutes to under 2 minutes (about 90% reduction). Demo shows segment-level quality and pipeline selection logic that runs at scale with minimal encoder overhead, integrates without DRM key sharing, and is already deployed in production.
AI-written summary · about
related talks
26mObjectionable Uses of Objective Quality Metrics
17mContent Aware Encoding for low latency live streaming encoders using deep learning
18mBlind Quality Assessment for Real-Time Visual Communications
03mJust realized we had a “Time Machine” after Demuxed’18 day 1
27mHighly-Efficient SVT-AV1-based Solutions for VOD Applications
22mThe User (has) Generated Content... so what now?
10mMoving Live Video Quality Control from the Broadcast Facility to the Living Room
19mRethinking the Video Transcoding Pipeline: How Pre-Filtering Holds the Key to Efficiency and Quality