Overview
Explore a conference talk on Swift, an adaptive video streaming system that leverages layered neural codecs. Dive into the innovative approach of using deep learning techniques for layered video coding, overcoming traditional limitations. Learn about Swift's key components: a layered encoder that efficiently encodes video frames, a decoder capable of fusing and decoding multiple layers simultaneously, and an adaptive bit rate protocol. Understand how Swift integrates seamlessly into existing streaming ecosystems and offers significant advantages over conventional video streaming systems. Gain insights into the system's performance comparisons and its potential impact on the future of video streaming technology.
Syllabus
Intro
Outline
Video Streaming Today
Limitations of Today's Streaming Alg
Existing Alternative: Layered Video St
Swift: Layered Coding Using Neural
2 Layered Neural Video Decod
3 Streaming Protocol/Algorithm
Putting Everything Together: Sv
Swift's Performance Comparist
Summary
Taught by
USENIX