Fourdee
World Intelligence
I’m Aayush Bansal. I build metric 4D world models — learning to reconstruct, understand, and create the visual world.
Explore my researchMaking the invisible
possible.
My research explores task- and domain-agnostic exemplar representations learned without supervision.
Virtual fashion.
Real motion.
Virtual try-on that moves beyond a single frame.
The research archive
Google Scholar
A visual world.
An open mind.
I earned my PhD in Robotics at Carnegie Mellon University, studying how to learn the 4D audiovisual world from sparse, unconstrained real-world samples.
I’ve worked at Meta’s Reality Labs Research in Pittsburgh and served as Principal Research Scientist at SpreeAI, developing machine learning for virtual try-on. My work also connects research with creative production through collaborations with production houses.
Carnegie Mellon
PhD · Robotics
Meta Reality Labs
Researcher · Pittsburgh
SpreeAI
Principal Research Scientist
Ideas that earned
recognition.
Uber Presidential
Fellowship
Named a Presidential Fellow at Carnegie Mellon University.
2017–18Qualcomm
Fellowship
Qualcomm Innovation Fellowship.
2019Snap
Fellowship
Snap Research Fellowship.
EgoVis Distinguished Paper Award
EgoHumans · ICCV oral presentation
EgoVis 2022/2023
Best Paper Award Finalist
Shapes and Context · CVPR
Oral presentation
Outstanding Reviewer
CVPR × 2 · ECCV × 1 · NeurIPS × 1
Academic service
CVPR 2024, 2025, 2026, 2027
AAAI 2026, 2027 · ECCV 2026
WACV 2027 · ICVGIP 2025
CVPR · ECCV · ICCV · NeurIPS · ICLR · ICML · SIGGRAPH · TPAMI · IJCV · TOG
Teaching at CMU
Geometry-based Methods
in Computer Vision
16-822 · Teaching Assistant
With Prof. Martial Hebert
Computer Vision
16-720 · Teaching Assistant
With Prof. Srinivasa Narasimhan
Research that
makes headlines.
Stories about 4D reconstruction, visual synthesis, and the possibilities they open up.
4D WORLD RECONSTRUCTIONNew System Combines Smartphone Videos To Create 4D Visualizations
Turning independently captured phone videos into scenes that can be explored from different viewpoints.
Read the storyThis tech (scarily) lets video change reality
Coverage of automated video transformations and their creative potential.
Read articleBeyond Deep Fakes: Transforming Video Content Into Another Video’s Style, Automatically
How temporal information helps Recycle-GAN learn transformations across video domains.
Read articleDon’t be fooled — this automated system sneakily manipulates video content
An interview about the ideas behind Recycle-GAN and its applications for artists and researchers.
Read article