Vision Systems and Intelligence LabVSI Lab

Reliable decisions from limited visual evidence.

The Vision Systems and Intelligence Laboratory at the University of Arizona builds vision systems that make reliable decisions with limited information and limited compute. We study how machines use images, video, and multimodal data to perceive, reason, and act, with applications in surgical and ophthalmic imaging, robotics, and intelligent sensing.

Join the lab. Prospective Ph.D., M.S., and undergraduate researchers and visiting scholars are welcome to email a CV and a short note on their research interests. eungjoolee@arizona.edu

Principal Investigator

Eungjoo Lee, Ph.D.

Assistant Professor, School of Electrical, Computing, and Software Engineering
Division Chief of Ophthalmic AI, Department of Ophthalmology and Vision Science
Director, Vision Systems and Intelligence Laboratory

Eungjoo Lee is an Assistant Professor in the School of Electrical, Computing, and Software Engineering at the University of Arizona and Division Chief of Ophthalmic AI in the Department of Ophthalmology and Vision Science. In summer 2026 he was a Visiting Researcher at the Schepens Eye Research Institute of Massachusetts Eye and Ear, Harvard Medical School. Before joining Arizona he was a postdoctoral research fellow in the Department of Radiology at Massachusetts General Hospital and Harvard Medical School, affiliated with the Martinos Center for Biomedical Imaging.

He received his Ph.D. in Electrical and Computer Engineering from the University of Maryland, College Park, and was a research intern at the U.S. Army Research Laboratory. He is affiliated with the Arizona Center for Telemedicine and Digital Health, the BIO5 Institute, the University of Arizona Cancer Center, and the Center for Semiconductor Manufacturing.

Google ScholarORCIDeungjoolee@arizona.edu

Eungjoo Lee, Ph.D.

Academic and Professional Service

Editorial

  • Scientific ReportsEditorial Board Member, Nature Portfolio2023–
  • Journal of Signal Processing SystemsAssociate Editor, Springer2023–
  • IEIE Transactions on Smart Processing and ComputingAssociate Editor2026–

Grant and fellowship review

  • National Institutes of HealthEarly Career Reviewer, Center for Scientific Review, Special Emphasis Panel on Biodata Management and Computational Modeling2026
  • Mitacs AccelerateResearch Proposal Reviewer2025–2026
  • National Science FoundationCSGrad4US Fellowship Reviewer2025

Advisory roles

  • NSF IUSE: Ethical Thinking and Case Analysis in AI EducationExternal Advisory Board Member2025–2028
  • Brain and Body Imaging Center, University of ArizonaScientific Advisory Committee2023–

Honors

  • Ralph E. Powe Junior Faculty Enhancement AwardOak Ridge Associated Universities2025
  • NSF I-CorpsEntrepreneurial and Technical Lead, Desert and Pacific Region Hub2024
  • NIJ Recidivism Forecasting ChallengeThird Place, U.S. Department of Justice2021

Conference organization

  • IEEE ISBITechnical Program Committee2026–2027
  • AAAIProgram Committee2025–2027
  • SPIE Defense + SecurityProgram Committee2025–2027
  • CVPRProgram Committee2026
  • Asilomar Conference on Signals, Systems, and ComputersSpecial Session Organizer2026
  • Hamlyn Symposium on Medical RoboticsWorkshop Organizer2025
  • IEEE-EMBS Body Sensor NetworksWorkshop Organizer and Technical Program Committee2025
  • IEEE ISBISession Chair2025
  • European Conference on Artificial IntelligenceProgram Committee2024–2025
  • SPIE Real-Time Image Processing and Deep LearningSession Chair2024

Reviewer

JournalsNature Machine Intelligence, Medical Image Analysis, Pattern Recognition, IEEE TPAMI, IEEE TMI, IEEE JBHI, IEEE TMC, IEEE TC

ConferencesCVPR, NeurIPS, ICML, ICLR, ECCV, MICCAI, AAAI, AISTATS, IPCAI, ISBI, ICIP, ICPR, ICASSP, WACV, ACCV

Current Research

Image, video, audio, and text inputs with the parts the model attends to for a text query highlighted; a decision map that refers, keeps routine care, or defers to a clinician by predicted risk and uncertainty; and a real robot that observes a covered container, removes the cover, re-observes, and grasps the target

Multimodal Perception and Decision Systems

Multimodal models that combine images, video, audio, and text, and decide when the visual evidence is sufficient to act. Robots look again before acting, and clinical models defer when uncertain.

A fundus photograph with the five-level diabetic retinopathy severity scale, retinal layers in OCT, brain regions on an MNI template linked as a graph with the matching connectivity matrix, human pose keypoints across video frames, and an endoscope camera trajectory

Medical Image and Video Understanding

Ophthalmic imaging, surgical video, human movement, and brain connectivity. We interpret retinal scans, estimate camera motion inside the body, track human pose over time, and build tools that assist clinicians during a procedure.

Two routes to compact models (a multi-head attention teacher distilled into a single-head student that fits a real-time budget, and pruning background tokens of a photograph of a statue so the model processes fewer tokens); a shift in feature space between training and deployment data; and a model that adapts on unlabeled images taken in clear, foggy, and rainy weather

Reliable and Efficient AI Systems

Models that hold up under distribution shift and still run on the hardware available at the bedside or in the field. We study test-time adaptation and knowledge distillation toward this goal.

More on our research