What to do after per-title encoding
summary
Per-title becomes a pipeline: build convex hulls with VMAF, target VMAF 94 for the top rendition, and train classifiers plus a convolutional LSTM to predict optimal resolution/bitrate pairs; deploy to production and handle live by per-segment classification and per-shot per-segment switching; choose midladder rates with reinforcement learning (tested with A3C and PPO), show simulated QoE gains over fixed ladders, and remain codec-neutral with an eye to AV1 and learned super‑resolution.
AI-written summary · about
chapters (4)
related talks
17mContent Aware Encoding for low latency live streaming encoders using deep learning
22mAI Enhanced GPU Video Coding: Achieving Joint High Compression Efficiency and Throughput
12mMeasuring live video quality with minimal complexity, now available for everyone!
30mMLDS 2019: Anubhav Shrivastava of Voot at Machine Learning Developers Summit 2019
18mCoVA: Exploiting Compressed-Domain Analysis to Accelerate Video Analytics
33mArchitecting Scalable Multi-Modal Video Search
32mEverything you wanted to know about VLMs but were afraid to ask
33mMeasuring Perceptual Video Quality at Scale