Brown CS News

Often described as the premiere conference of the computer vision community, Computer Vision and Pattern Recognition (CVPR, held in June this year in Nashville, Tennessee) is large and highly competitive. In 2025, it received 13,008 submissions for publication from over 30,000 international authors. Through the competition of peer review, most of the accepted works (22%) were presented as posters, with less than 100 selected for oral presentation (0.7%) and only 15 shortlisted as a best paper award candidate (0.1%).

This year, research from Brown Visual Computing (BVC) student Yiqing Liang and her collaborators at NVIDIA was recognized as a best paper award candidate with oral presentation. Yiqing’s work uses AI to predict the 3D position and motion of every pixel given just two frames of a video, which is a foundational capability underpinning many applications across robotics and autonomous vehicles, or even in media for video editing and visual effects (VFX). The big challenge is making it work across different scenes. Yiqing says that “it’s really cool how general it is! We’ve tested it on out-of-domain datasets – real-world, high-motion scenes – and it still works!”

Another oral from BVC Master’s student Runfeng Li also investigates 3D position and motion estimation, but this time using time of flight imaging, similar to Lidar. This sends additional light into the scene to accurately reconstruct fast motion like swinging baseball bats. To move towards object-specific models, Rao Fu’s research on detailed 3D hand tracking introduces a large dataset for bimanual interaction with tools, and Arthur Chen and Chaerin Min’s research improves synthesis of hand images – something that AI currently struggles with. Both were presented as poster highlights (3%). Aashish Rai’s research creates a new way to compactly represent 3D scenes (UVGS), but BVC isn’t all about estimating 3D and motion: Aditya Ganeshan’s work on patterns such as those in textiles or print design creates new ways to intuitively edit those patterns by analogy.

BVC alums are also at the top of their game. Research from BVC undergraduate and Master’s alum Benjamin Attal (CMU PhD) and colleagues was awarded the best student paper for their work on imaging light in flight and modeling its interreflections. Ben follows BVC undergraduate alum David Charatan (MIT PhD) who, with his colleagues, was awarded a best paper honorable mention prize at CVPR 2024 for their work on learning to predict 3D scenes efficiently and effectively from two images. BVC undergraduate alum Adam Pikielny, in his role as a Research Engineer at Adobe, also presented research as an oral on how to separate reflections in glass from the scene behind to improve computational photography editing.

Incoming Brown PhD student Xiaoyi (Jason) Liu also presented his work on image super-resolution using diffusion models and operator learning as an oral: “I chose Brown Visual Computing to study because of their innovative atmosphere, cutting-edge research, and supportive community.”

Of course, the professors are no slouches either. Brown CS faculty member Chen Sun, with his colleagues at Google, including Brown CS alum Deqing Sun, presented research as an oral on how to use motion prompts in AI to animate images to look like videos. But the field moves fast! Its follow-up work, Force Prompting, authored by BVC PhD student Nate Gillman and currently in review, was just featured on local news affiliate NBC 10 WJAR for its ‘wow’ effect.

Referenced Brown CS Works

Zero-shot Monocular Scene Flow Estimation In The Wild
https://research.nvidia.com/labs/lpr/zero_msf/ 
Yiqing Liang, Abhishek Badki*, Hang Su*, James Tompkin, Orazio Gallo (*equal advising)
CVPR 2025 Best Paper Award Candidate

Time of the Flight of the Gaussians: Optimizing Depth Indirectly in Dynamic Radiance Fields
https://visual.cs.brown.edu/gftorf 
Runfeng Li, Mikhail Okunev, Zixuan Guo, Anh Duong, Christian Richardt, Matthew O’Toole, James Tompkin
CVPR 2025 Oral

GigaHands: A Massive Annotated Dataset of Bimanual Hand Activities
https://ivl.cs.brown.edu/research/gigahands.html 
Rao Fu, Dingxi Zhang, Alex Jiang, Wanjia Fu, Austin Funk, Daniel Ritchie, Srinath Sridhar
CVPR 2025 Highlight

FoundHand: Large-Scale Domain-Specific Learning for Controllable Hand Image Generation
https://ivl.cs.brown.edu/research/foundhand.html 
Kefan (Arthur) Chen, Chaerin Min, Linguang Zhang, Shreyas Hampali, Cem Keskin, Srinath Sridhar
CVPR 2025 Highlight

UVGS: Reimagining Unstructured 3D Gaussian Splatting using UV Mapping
https://aashishrai3799.github.io/uvgs/ 
Aashish Rai, Dilin Wang, Mihir Jain, Nikolaos Sarafianos, Arthur Chen, Srinath Sridhar, Aayush Prakash
CVPR 2025

Pattern Analogies: Learning to Perform Programmatic Image Edits by Analogy
https://bardofcodes.github.io/patterns/ 
Aditya Ganeshan, Thibault Groueix, Paul Guerrero, Radomír Měch, Matthew Fisher, Daniel Ritchie
CVPR 2025

Motion Prompting: Controlling Video Generation with Motion Trajectories https://motion-prompting.github.io/
Daniel Geng, Charles Herrmann, Junhwa Hur, Forrester Cole, Serena Zhang, Tobias Pfaff, Tatiana Lopez-Guevara, Carl Doersch, Yusuf Aytar, Michael Rubinstein, Chen Sun, Oliver Wang, Andrew Owens, Deqing Sun
CVPR 2025 Oral

Force Prompting: Video Generation Models Can Learn and Generalize Physics-based Control Signals
https://force-prompting.github.io/ 
Nate Gillman, Charles Herrmann, Michael Freeman, Daksh Aggarwal, Evan Luo, Deqing Sun, Chen Sun

For more information, click the link that follows to contact Brown CS Communications Manager Jesse C. Polhemus.