Towards Robust and Practical Neural Video Conferencing

Speaker: Vibhaalakshmi Sivaraman , MIT CSAIL

Date: Monday, October 16, 2023

Time: 10:00 AM to 11:00 AM Note: all times are in the Eastern Time Zone

Public: Yes

Location: Seminar Room D463 (Star)

Event Type: Thesis Defense

Room Description:

Host: Mohammad Alizadeh

Contact: Vibhaalakshmi Sivaraman, vibhaa@csail.mit.edu

Relevant URL:

Speaker URL: https://people.csail.mit.edu/vibhaa/

Speaker Photo:
None

Reminders to: seminars@csail.mit.edu, g9@csail.mit.edu

Reminder Subject: TALK: Monday 10-16-2023 Thesis Defense: "Towards Robust and Practical Neural Video Conferencing"

Abstract: Video conferencing systems suffer from poor user experience when network conditions deteriorate because current video codecs simply cannot operate at extremely low bitrates. Recently, several neural alternatives have been proposed that reconstruct talking head videos at very low bitrates using sparse representations of each frame such as facial landmark information. However, these approaches produce poor reconstructions in scenarios with major movement or occlusions over the course of a call, and do not scale to higher resolutions. This thesis first presents Gemino, a new neural compression system for video conferencing based on a novel high-frequency-conditional super-resolution pipeline. Gemino upsamples a very low-resolution version of each target frame while enhancing high-frequency details (e.g., skin texture, hair, etc.) based on information extracted from a single high-resolution reference image. We use a multi-scale architecture that runs different components of the model at different resolutions, allowing it to scale to resolutions comparable to 720p, and we personalize the model to learn specific details of each person, achieving much better fidelity at low bitrates. We implement Gemino atop aiortc, an open-source Python implementation of WebRTC, and show that it operates on 1024x1024 videos in real-time on a Titan X GPU, and achieves 2.2–5x lower bitrate than traditional video codecs for the same perceptual quality. The thesis will also cover alternate designs for neural video conferencing solutions that leverage attention techniques instead of optical flow, along with neural compression techniques for loss-resilience.

Research Areas:
Systems & Networking

Impact Areas:

This event is not part of a series.

Created by Vibhaalakshmi Sivaraman Email at Thursday, September 28, 2023 at 3:15 PM.