Tai-Sing Lee

Also known as Tai Sing Lee

Tai-Sing Lee is a professor of computer science at Carnegie Mellon University whose work links visual perception, neural computation and computer vision. He trained at Harvard College, completed doctoral work at Harvard and the Harvard-MIT Division of Health Science and Technology in 1993, and joined Carnegie Mellon in 1996. With David Mumford, he published a Bayesian account of visual processing in 2003; he received the ICCV Helmholtz Prize in 2013. He led a MICrONS team in 2016 to study brain circuits for machine learning and coauthored a 2025 method that uses boundary and surface representations to improve visual learning.

Now at: Carnegie Mellon University
Works: 102
Citations: 6,138
h-index: 28
Badges: Loyal

Publication numbers from OpenAlex, as of 27 October 2026.

Timeline

1986–1996

Harvard degrees and postdoctoral research

He earns an SB in engineering physics at Harvard College in 1986. In 1993, he completes doctoral work in engineering sciences at Harvard University and in medical physics and engineering at the Harvard-MIT Division of Health Science and Technology. From 1993 to 1996, he works as a postdoc in biomedical physics and neurophysiology at Harvard University (DAS) and MIT (BCS).
1996

Joining Carnegie Mellon

He joins Carnegie Mellon University in 1996.
1996

A visual cortex grant

In 1996, he leads Carnegie Mellon’s project The Role of the Primary Visual Cortex in High Level Perceptual Computation. The James S. McDonnell Foundation awards the university $105,000 for the three-year project.
2003

Bayesian inference in visual cortex

With David Mumford, he publishes Hierarchical Bayesian Inference in the Visual Cortex in 2003. The paper proposes that recurrent connections combine contextual expectations with sensory observations to perform probabilistic inference across the visual hierarchy.
2013

The ICCV Helmholtz Prize

He receives the ICCV Helmholtz Prize in 2013.
January 2016

Leading a MICrONS team

In January 2016, he leads a Carnegie Mellon MICrONS team as principal investigator, with Sandra Kuhlman and Johns Hopkins University’s Alan Yuille as co-principal investigators. The team partners with a Harvard-led consortium to study neural circuits and develop computational models and machine learning algorithms.
2025

Perception before contrastive learning

In 2025, he coauthors Perceptual Inductive Bias Is What You Need Before Contrastive Learning at CVPR. The method first learns boundary and surface representations, then trains for object semantics; the experiments show twice as fast convergence on ResNet18 and improvements in segmentation, depth estimation and object recognition.
As of October 2026

Professor at Carnegie Mellon

In October 2026, he is a professor in Carnegie Mellon’s Computer Science Department, working on computational neuroscience and visual perception. His laboratory uses computational, mathematical and neurophysiological methods to study neural computation and improve computer vision.