UIST 2026 · PosterFirst author
CoGaze: Webcam-Based Gaze Sharing
Gaze sharing from the webcam you already have
Tokyo, Japan
Takuma Masuda
I build technology that helps people take in more of the world, and more of each other.
I am an undergraduate at Nojima Laboratory, University of Electro-Communications, working on human augmentation: technology that extends how people perceive, act and understand, alongside the people around them rather than on their behalf. So far that has meant shared gaze (CoGaze, at NUS) and multisensory VR.
Published
UIST 2026
First-author poster, as an undergraduate: CoGaze.
Abroad
Augmented Human Lab, National University of Singapore. From first prototype to a UIST poster in one internship, in English.
Next
2027
Graduating in March. Looking for a master’s or PhD lab, or an RA / RE role.
Recent work
Things I built. What they do, before what they are.
IVRC 2026Team Eaterminator
Meal Be Back
Chewing, played backwards
Side projectJava
Planarity, graph puzzle game
Untangle the lines until none of them cross
seichi.clickMinecraft · Scala
Open source · rolled backScala
Minecraft plugin: rolled back
What broke when I gave back in code
Research
Research interests
In short
People who can take in more of the world, and more of each other.
Fields
Augmented Human
Assistive Augmentation
Human-Computer Interaction
Virtual Reality
Cross / Extended Reality
Social acceleration by wearables
Assist agent design
01
Could you borrow a sense from someone else?
For example
A remote expert's gaze landing on the part you are holding.
A grandparent seeing exactly what a child is pointing at across the room.
A view someone cannot see, felt rather than described.
Shared gaze works in MR when both people wear the same headset. CoGaze tried something close to the first across devices that do not match, with a webcam on one side.
02
Could the body carry more?
For example
Wearables close to the body, from the jaw to the eyes, that add a sensation or a channel you did not have.
VR bodies that change what you feel you are able to do.
A glance or a hesitation passed between people fast enough to speed up how a group comes together.
03
Could you understand for yourself, faster?
For example
An assistant that points to what matters instead of summarising it.
One that knows when to step back.
AI that grows a person's own judgement and fits the relationships they already have, rather than standing in for them.
Assist agent designHCI
Under all three: design for devices that do not match and signals that are sometimes wrong, and study how people repair them. Hardware will keep changing; that mismatch will not go away.
What I am actually after
Lend me your eyes.
01 · The pipe
The world is not simply out there. Your brain builds it out of whatever reaches your senses. For a lot of people, a lot of that now arrives through a screen: one small rectangle, read with the eyes, a line at a time. That is a narrow pipe for a whole world. Anyone who uses one runs into its limits: you get tired, you skim, you miss things. For some people the pipe is closed most of the way: older people, people with a disability, people living in a language they did not grow up in. Same pipe, different width.
02 · The shortcut
The answer on offer is to let a model read it for you. It works, but you stop looking. For a chore that is fine. For something you should judge yourself, you are acting on an answer whose reasons you never saw. And taste, values, what matters to you are things you can only get by living them.
03 · Make it visible
Both problems are the same problem: the person is not seeing anything. So do not answer for people. Make the thing visible: put it where eyes and body can take it in, with nothing to decode. And when it really does take an expert, borrow their eyes, not their answer: let them show you what they are looking at.
04 · A first step
CoGaze asked a small version of that question: can one person hand their attention to another using nothing but the webcam they already own? It can. Shared gaze was already known to work in MR when everyone wears the same headset. The new part was sharing it across devices that do not match.
If each of us can build values of our own, and they still fit the people around us, the world opens up to everyone.
Now
Short updates, newest first.
Next
Sharing attention across different devices
Shared gaze already works in MR, when everyone wears the same headset.
Real teams mix devices: a laptop on one side, smart glasses on the other.
How should attention be shown when two people sense and see differently?
CoGaze was one such pairing. The next step is taking shape at Nojima Laboratory.
{{ item.date }}
Experience
2026 – now
Undergraduate Researcher
Nojima Laboratory, University of Electro-Communications
2025
Research Intern
Augmented Human Lab, National University of Singapore
2025 – now
Teaching Assistant, Basic Programming
UEC high-school partnership programme
Education
2023 – 2027
B.Eng., Informatics & Engineering
University of Electro-Communications, Tokyo
2027
Minor, English for Science & Technology
UEC Minor Program · Open Badge on completion
2025-10
UEC Data Scientist Training Program
Applied foundational level · Verify Open Badge ↗
Support · 2025
UEC Foundation Scholarship
Supported my research internship at NUS, where CoGaze was built and tested. Thank you. Support UEC students ↗
Toolkit
Programming
PythonJavaC#JavaScriptHTMLCScalaRubyRGLSL / HLSLProcessingLaTeX
XR & sensing
UnityMeta QuestVR / AR / MREye gaze trackingMediaPipeWebcam CVOpenCVTobii / Pupil LabsOpenXRSteamVRBlenderPhotogrammetry / 3D scan
ML & data
pandas / NumPyJupyterscikit-learnPyTorchLLM APIs & promptingKaggle
Hardware
Arduino / ESP32Soldering & circuits
Dev tools
Git / GitHubDockerWeb / GitHub Pages
CS & math
AlgorithmsData structuresOperating systems (C)Networks (C)Discrete mathematicsFormal language theoryProbability & statisticsLinear algebraSignal processingCalculus & analysisComplex analysisOperations research
Languages
Japanese (native)English (fluent)
Hobby
Ericksonian hypnosisPossibility therapyNLP
Read for its own sake, and unexpectedly useful whenever I have to run a session with a person in it.
Get in touch.
Graduating March 2027. Open to graduate research positions, RA and RE roles, internships and collaborations in HCI, VR and human augmentation, and happy to hear about anything adjacent.