I am a researcher at the intersection of computer graphics, vision, and networked systems. I currently work at NVIDIA Research in the AMRI research group. My focus is practical and resilient neural video and graphics, particularly compressed representations for streaming media. My research efforts span 3D and 4D representations, visual perception, and systems and networking.

Before joining NVIDIA, I ran a startup commercializing neural video compression. I received my PhD from the Allen School of Computer Science at the University of Washington. My dissertation proposed perceptual optimizations for visual computing hardware accelerators, storage systems, and data management systems. I received my bachelors degree at Columbia University.

Interns: I am always looking for motivated interns. If you are a PhD student interested in computer vision/graphics/generative AI for compressed 3D and 4D representations, particularly for use in interaction or multimodal learning systems, please send me an email.

Recent News (see all →)

I will be attending NeurIPS in Vancouver to present our paper QUEEN.
December 2024
Yihua presented GRACE at NSDI 2024!
December 2024
We presented AI-Mediated 3D Videoconferencing, a SIGGRAPH Emerging Technologies Demo. Thanks to everyone who came to check it out!
August 2023
Served on Program Committees for PACT, IISWC, SIGCOMM EMS, and ISCA ASSYST, and I externally reviewed for SIGGRAPH Asia.
July 2023

Publications

QUEEN: QUantized Efficient ENcoding for Streaming Free-viewpoint Videos.
Sharath Girish, Tianye Li, Amrita Mazumdar, Abhinav Shrivastava, David Luebke, Shalini De Mello.
In NeurIPS, 2024.
We present an efficient representation for streamable free-viewpoint videos with dynamic Gaussians. Our method QUEEN is able to capture dynamic scenes at high visual quality and reduce the model size to just 0.7 MB per frame while training in under 5 seconds and rendering at ∼350 FPS.

GRACE: Loss-Resilient Real-Time Video through Neural Codecs.
Yihua Cheng, Ziyi Zhang, Hanchen Li, Anton Arapin, Yue Zhang, Qizheng Zhang, Yuhan Liu, Xu Zhang, Francis Y. Yan, Amrita Mazumdar, Nick Feamster, Junchen Jiang.
In NSDI, 2024.
We present a loss-resilient real-time video system called GRACE, which preserves the user's quality of experience (QoE) across a wide range of packet losses through an autoencoder-based neural video codec.

Online Overexposed Pixels Hallucination in Videos with Adaptive Reference Frame Selection.
Yazhou Xing, Amrita Mazumdar, Anjul Patney, Chao Liu, Hongxu Yin, Qifeng Chen, Shalini De Mello, Jan Kautz, Iuri Frosio.
In arXiV, 2023.
We present a learning-based system to hallucinate HDR content in overexposed video sequences, without resorting to complex acquisition mechanisms like alternating exposures or costly processing typical of HDR imaging.

AI-Mediated 3D Video Conferencing.
Michael Stengel, Koki Nagano, Chao Liu, Matthew Chan, Alex Trevithick, Shalini De Mello, Jonghyun Kim, David Luebke, Amrita Mazumdar, Shengze Wang, Mayoore Jaiswal.
In SIGGRAPH Emerging Technologies, 2023.
We present an AI-mediated 3D video conferencing system that can reconstruct and autostereoscopically display a life-sized talking head using consumer-grade compute resources and minimal capture equipment.

VSS: A Storage System for Video Analytics.
Brandon Haynes, Maureen Daum, Dong He, Amrita Mazumdar, Magda Balazinska, Alvin Cheung, Luis Ceze.
In SIGMOD, 2021.
A video storage system for video data management that enables fine-grained access to video content, caching, and redundancy elimination for overlapping field-of-view.

TASM: A Tile-Based Storage Manager for Video Analytics.
Maureen Daum, Brandon Haynes, Dong He, Amrita Mazumdar, Magda Balazinska, Alvin Cheung.
In IEEE International Conference on Data Engineering, 2021.
A tile-based storage manager enabling spatial random access to encoded videos for analytics workloads.

VisualWorldDB: A DBMS for the Visual World.
Brandon Haynes, Maureen Daum, Amrita Mazumdar, Magda Balazinska, Luis Ceze, Alvin Cheung.
In Conference on Innovative Data Systems Research (CIDR), 2020.
paper (pdf), bibtex

A vision and initial architecture for a new type of database system optimized for large-scale multicamera applications.

Vignette: Perceptual Compression for Video Storage and Processing Systems.
Amrita Mazumdar, Brandon Haynes, Magda Balazinska, Luis Ceze, Alvin Cheung, Mark Oskin.
In ACM Symposium on Cloud Computing (SoCC), 2019.
paper (pdf), slides (pdf), more recent slides (pdf), bibtex, SoCC Best Poster Award Winner

A system that integrates machine learning-improved compression with cloud video storage and distribution, compatible with modern codecs and hardware accelerators.

Visual Road: A Video Data Management Benchmark.
Brandon Haynes, Amrita Mazumdar, Magda Balazinska, Luis Ceze, Alvin Cheung.
In SIGMOD, 2019.
paper (pdf), bibtex

A scalable analytics benchmark suite and video generator for video databases.

LightDB: A DBMS for Virtual Reality.
Brandon Haynes, Amrita Mazumdar, Armin Alaghi, Magda Balazinska, Luis Ceze, Alvin Cheung.
In Proceedings of the VLDB Endowment (PVLDB) 11(10), 2018.
paper (pdf), bibtex, code (github)

A database management system designed for multi-dimensional video, like 360-degree and light field videos.

Application Codesign of Near-Data Processing for Similarity Search.
Vincent T. Lee, Amrita Mazumdar, Carlo C. Del Mundo, Armin Alaghi, Luis Ceze, Mark Oskin.
In IEEE International Parallel and Distributed Processing Symposium (IPDPS), 2018.
paper (pdf), bibtex

A k-nearest neighbors hardware accelerator using processing-in-memory, for content-based image retrieval.

A Hardware-Friendly Bilateral Solver for Real-Time Virtual Reality Video.
Amrita Mazumdar, Armin Alaghi, Jonathan T. Barron, David Gallup, Luis Ceze, Mark Oskin, Steven M. Seitz.
In High Performance Graphics (HPG), 2017.
paper (pdf), slides (pdf), bibtex, code (github), blog post

A hardware-software codesign approach to accelerate a 16-camera VR video pipeline for real-time performance.

Exploring Computation-Communication Tradeoffs in Camera Systems.
Amrita Mazumdar, Armin Alaghi, Thierry Moreau, Sung Min Kim, Meghan Cowan, Luis Ceze, Mark Oskin, Visvesh Sathe.
In IEEE International Symposium on Workload Characterization (IISWC), 2017.
paper (pdf), slides (pdf), bibtex

A data movement characterization for resource-constrained vision and VR camera hardware.

Principles and Techniques of Schlieren Imaging Systems.
Amrita Mazumdar.
In Columbia University Computer Science Technical Reports, 2013. , bibtex
A survey paper on modern Schlieren optics systems.