Physics-Informed Vision and Imaging

Aviad Levis
Kevin McKinnon
Rohan Dahale
Alexandra Worms
Ali SaraerToosi
David Bromley
Hannah Collier
Isabelle Laing
Len Luong
Mickell Als
Peter Phan
Renbo Tu
Thanosan Prathifkumar
Our members

Physics-Informed Vision and Imaging is a research group focused on advancing the science of imaging through novel sensors, physics-aware computation, and machine learning. We push the limits of what can be seen — revealing new insights about our world and the cosmos.

Research Focus

Observational Tools

Instruments

Imaging systems across radio, optical, and time-resolved sensing.
Algorithms & Theory

Computation

Physics-informed machine learning and neural rendering for inverse problems, differentiable physics, dynamic inference, and scientific simulation.
Application Domains

Science

From particles to planets to black holes — discovery across all scales.

Recent Publications

View all publications
Video reconstruction of variable VLBI observations with neural fields
Nature 2026 Foschi, Zhao et al.
NeuralDMD: Interpretable Neural Representation of Dynamics from Sparse and Noisy Measurements
ECCV 2026 SaraerToosi, Tu et al.
Telemetry is a Sensor: Opportunistic Wavefront Estimation for the James Webb Space Telescope
ICCP 2026 Buzi et al.
A cloud’s droplet size—and its internal structure—from monochrome multi-angle imaging
IEEE Transactions on Geoscience and Remote Sensing 2026 Davis et al.
BHCast: Unlocking Black Hole Plasma Dynamics from a Single Blurry Image with Long-Term Forecasting
CVPR 2026 Tu et al.
Dynamic Black-hole Emission Tomography with Physics-informed Neural Fields
CVPR 2026 Feng et al.

News & Highlights

Pinned highlights

Pinned

We are looking for students interested in working this summer on intensity interferometry for high-resolution astronomy. The project will explore SPAD-based photon timing, telescope arrays, and computational imaging methods for studying compact astronomical sources. To express interest, please complete the Google Form.

Pinned

Hannah Collier has been awarded a Schmidt AI in Science Fellowship and will join the group in Summer 2026. Hannah will work on computer vision and 3D reconstruction methods for solar observations and heliophysics.

Date News

Check out our latest work, “Astronomy’s Summer Blockbuster”: an AI-constructed movie of a blazar jet. Jointly led by Marianna Foschi, Brandon Zhao, and Antonio Fuentes, this work was published in Nature.

New paper out in IEEE Transactions on Geoscience and Remote Sensing: “A cloud’s droplet size—and its internal structure—from monochrome multi-angle imaging”.

A new paper on wavefront error estimation from JWST telemetry, led by Lahav Buzi (Technion), was presented at ICCP 2026.

CloudCT-Precursor is set to launch on SpaceX’s Transporter-17 mission. This mission builds on our cloud tomography retrieval work [1], [2], [3], [4].

NeuralDMD: Interpretable Neural Representation of Dynamics from Sparse and Noisy Measurements has been accepted to ECCV 2026. Congratulations to Ali SaraerToosi!

Two papers BHCast: Unlocking Black Hole Plasma Dynamics from a Single Blurry Image with Long-Term Forecasting and Dynamic Black-hole Emission Tomography with Physics-informed Neural Fields have been accepted to CVPR 2026. Congratulations to Renbo Tu and David Bromley!

Rohan Dahale received two awards from the Event Horizon Telescope Collaboration: the Outstanding PhD Thesis Award — for showing that M87’s ring ellipticity is governed by turbulent plasma dynamics and for establishing the validation framework that underpins the first robust videos of SgrA* — and the EHT Early-Career Award 2025 — for his work on investigating the origin of the ring ellipticity in the black hole images of M87*. Congratulations, Rohan!

Aviad Levis received $400K award from the CFI John R. Evans Leaders Fund (JELF) and Ontario Research Fund (ORF-RI) to lead a research program in Physics-constrained neural fields for scientific imaging. Read more from U of T: 55 U of T scholars receive research infrastructure support.

Supported by the Data Sciences Institute, Aviad Levis is co-leading Bridging the Gap: From Computational Physics, to Physics-informed Machine Learning, to Data-driven Scientific Discovery. The DSI Emerging Program unites four co-leads and 16 members across nine departments to connect first-principles simulations with data-driven discovery for high-impact domains—from sustainable design to astrophysics.

Rohan Dahale and Kevin McKinnon were selected for the 2025 cohort of the Schmidt AI in Science Fellowship. Their fellowships kick off this September, supporting bold AI-driven approaches to scientific discovery.

Aviad Levis received the 2024–25 Connaught New Researcher Award, supporting high-impact directions in physics-informed vision and imaging.

David Bromley was awarded the Ontario Graduate Scholarship (OGS) for 2025–2026, recognizing his outstanding research contributions.

U of T highlighted Aviad Levis in “In awarding Nobel Prize, committee cites collaborative black hole research including U of T computer scientist Aviad Levis,” recognizing his Event Horizon Telescope contributions to the discoveries celebrated by the Nobel Committee.

The Natural Sciences and Engineering Research Council of Canada awarded Aviad Levis an NSERC Discovery Grant to launch his research program in scientific imaging and physics-informed computation.