AI Enhanced GPU Video Coding: Achieving Joint High Compression Efficiency and Throughput

on YouTube ↗22m

" summary

Describes an approach that pairs Nvidia NVENC hardware encoding with ML-driven pre-analysis and pre-processing to improve compression efficiency without losing throughput. Techniques include scene and region classification, ROI detection (e.g., faces, jersey numbers), deblurring/denoising/sharpening/contrast pre-processing, and hierarchical pre-classification with fine-grained QP control; results on NVIDIA T4 show 15–20% bitrate savings and ~1–2 VMAF gain for UGC/PUGC while keeping parallel encoding throughput.

AI-written summary · about

" related talks

• 17mGemino: Practical and Robust Neural Compression for Video ConferencingMohammad Alizadeh, Pantea Karimi, Vibhaalakshmi Sivaraman, Vedantha Venkatapathy, Mehrdad Khani, Sadjad Fouladi, Frédo Durand, Vivienne Sze · 21st USENIX Symposium on Networked Systems Design and Implementation (NSDI '24)
• 34mHow NVIDIA is Advancing Video Curation with Generative AINiket Agarwal · Ray Summit 2024
• 24mASIC-RDX and Compute-Compression Efficiency Optimization for UGC Video ProcessingShankar Regunathan, Srinath Reddy · Video @Scale 2021
• 21mWhat to do after per-title encodingBen Dodson, Nick Chadwick · Demuxed 2018
• 17mContent Aware Encoding for low latency live streaming encoders using deep learningRamdas Satyan · Demuxed 2023
• 11mVirtues of Recycling in Multi-rate EncodingAlex Giladi · Demuxed 2018
• 35mLarge-Scale Video IntelligenceIrene Donato, Antonino Ingargiola · PyData Amsterdam 2025
• 18mCoVA: Exploiting Compressed-Domain Analysis to Accelerate Video AnalyticsJongse Park, Jinwoo Hwang, Minsu Kim, Daeun Kim, Seungho Nam, Yoonsung Kim, Dohee Kim, Hardik Sharma, Minsu Kim, Daeun Kim, Yoonsung Kim, Dohee Kim · USENIX ATC '22