Phat Nguyen (Peter)

I'm currently a visiting researcher at the MIT Distributed Robotics Lab. I completed my Bachelor's in Computer Science with a minor in Mathematics at the University of Massachusetts Amherst, where I was a research assistant at the Computer Vision Lab under Prof. Erik Learned-Miller.

My research interest lies at the intersection of computer vision, machine learning, and robotics. My goal is to bridge the gap between high-level cognition and low-level planning and control, with applications for autonomous vehicles.

Email  /  Resume  /  LinkedIn  /  Github

peter nguyen's photo

My area of research interest is in autonomous vehicles, computer vision, machine learning, and robotics.

Robust Frame-to-Frame Camera Rotation Estimation in Crowded Scenes
Fabien Delattre, David Dirnfeld, Phat Nguyen, Stephen Scarano, Pedro Miraldo, Michael Jeffrey Jones, Erik Learned-Miller
International Conference on Computer Vision (ICCV), 2023
project page / poster / arXiv

We introduce a novel generalization of the Hough transform on SO3 to efficiently find the camera rotation most compatible with the optical flow. We also provide a new dataset of 17 highly dynamic video sequences called BUsy Street Scenes (BUSS).

Sept. 2023,

TA for CS326: Web Programming, UMass

May 2023,

TA for CS240: Reasoning Under Uncertainty, UMass

Jan. 2023,

TA for CS389: Machine Learning, UMass (Received distinguished TA award)

Activities and Leadership

Previously, I co-founded Stella Agritech to create technologies that enhance the efficiency of agriculture. I also led ISHCMC's sustainability program with Tanya Meftah, which was nominated by the European Chambers of Commerce for the Best Sustainable Business Initiative in Vietnam in 2019. I was also part of the UMass Rocket Team between 2021-2022.

UMass Aerospace and Rocket Team, NASA SLI 2022, Payload Software Engineer

Building the onboard flight inertial reference system (IRS) to continuously compute the launch vehicle's dead reckoning position in a GPS-denied environment with a 75 meter (250 ft) resolution


Some inspiration from Jon Barron.