Visual Physics @ CVPR 2020

Date: June 19, 2020

Click Here for the Zoom Webinar!


1:00 - 2:00PM (PT)

Opening Panel - literature will be surveyed and papers will be suggested for reading during the break

Bill Freeman, MIT

Achuta Kadambi, UCLA

Laura Waller, Berkeley

Ayan Chakrabarti, WUSTL

Katerina Fragkiadaki, CMU

2:00 - 3:30PM (PT)

Break - selected papers for reading are listed at the bottom of the webpage

3:30 - 4:30PM (PT)

Closing Panel - we will go deeper into literature after folks have had a chance to read, and suggest future trends for the community.


Abstract: The early days of computer vision used "physics-inspired algorithms" to detect contours, edges, faces, and other features. Over the past decade, these structured algorithms have been superseded by deep learning algorithms with superior performance. This tutorial discusses an increasingly popular class of hybrid methods that blend physics and learning. In a unique virtual format, we feature 5 faculty members who provide a joint opening panel and closing panel, with a flipped classroom style reading group in the middle break. A happy hour will conclude the tutorial.

Achuta Kadambi

Achuta Kadambi
Assistant Professor at UCLA

Bill Freeman

Bill Freeman
Thomas and Gerd Perkins Professor at MIT


Laura Waller

Laura Waller
Professor at UC Berkeley

Katerina Fragkiadaki

Katerina Fragkiadaki
Assistant Professor at CMU


Ayan Chakrabarti

Ayan Chakrabarti
Assistant Professor at WUSTL

Visual Physics: Discovering Physical Laws from Videos

Phase-Based Video Motion Processing

Turning Corners into Cameras: Principles and Methods

Physics-Based Learned Design: Optimized Coded-Illumination for Quantitative Phase Imaging

Learning Spatial Common Sense with Geometry-Aware Recurrent Networks

Training Image Estimators without Image Ground-Truth

Learning visual predictive models of physics for playing billiard

Use the Force, Luke! Learning to Predict Physical Forces by Simulating Effects

A theory of fermat paths for non-line-of-sight shape reconstruction

TossingBot: Learning to Throw Arbitrary Objects with Residual Physics

Interaction networks for learning about objects, relations and physics

Learning Spatial Common Sense with Geometry-Aware Recurrent Networks

Learning from Unlabelled Videos Using Contrastive Predictive Neural 3D Mapping

Auto-Tuning Structured Light by Optical Stochastic Gradient Descent

How to Do Physics-based Learning

Integrating Physics-Based Modeling with Machine Learning: A Survey

Heavy Rain Image Restoration: Integrating Physics Model and Conditional Adversarial Learning

AI Feynman: A physics-inspired method for symbolic regression

Integration of Neural Network-Based Symbolic Regression in Deep Learning for Scientific Discovery

Discovering Symbolic Models from Deep Learning with Inductive Biases

Deep Shape from Polarization

Learning Topology From Synthetic Data for Unsupervised Depth Completion

Blending Diverse Physicsl Priors With Neural Networks

Learning From Synthetic Data: Addressing Domain Shift for Semantic Segmentation


Responsive image

Webpage design courtesy of J-F Lalonde and M Gupta