Nhut Le

Nhut Le

CS PhD student, Lehigh University

I am a second-year PhD student in Computer Science at Lehigh University, advised by Prof. Maryam Rahnemoonfar. My current research interest lies in deep learning and computer vision, and their application in 3D reconstruction, 3D rendering and 3D semantic segmentation.

Previously, I had been working as a 3D computer vision engineer at Sturfee, Inc. for three years. My work focused on 3D reconstruction and 3D localization to improve the performance and accuracy of Visual Positioning System (VPS). I received my M.S. and B.S. in Computer Science at Portland State University.

Download resume

🔥 News

  • 2024.08.12:  First-day at Lehigh University as a PhD student.

Publications

Paper Teaser
Paper Teaser
Nhut Le, Maryam Rahnemoonfar
Geospatial Informatics XV, 2025
Paper Teaser
Nhut Le, Maryam Rahnemoonfar
the 2025 IEEE International Geoscience and Remote Sensing Symposium (IGARSS 2025)

Experience

  • 2024.08 - present,  Research Assistant - Bina Lab, Lehigh University, Bethlehem, PA
    • Research 3D methodologies and machine learning to support disaster-affected areas
  • 2020.10 - 2023.10,  Computer Vision Engineer - Sturfee, Inc., Remote
    • Developed, prototyped, and enhanced methods for 3D reconstruction and precise localization in the 3D prior map using Structure-from-Motion (SfM) and Multi-View Stereo (MVS) techniques, resulting in improving outdoor VPS and enabling indoor VPS
    • Implemented and optimized RGBD integration algorithms to reconstruct 3D point clouds and meshes, such as using TSDF volume integration and optimizing RGBD poses
    • Worked with deep learning networks for 3D semantic segmentation
    • Proficient in 3D manipulations such as mesh texturing, texture baking, bundle adjustment, and UV unwrapping

Educations

  • 2024.08 - present,  PhD in Computer Science, Lehigh University.
  • 2018.09 - 2020.08,  Master’s in Computer Science, Portland State University.
  • 2015.01 - 2018.06,  Bachelor in Computer Science, Portland State University.

Honors and Awards

  • 2015 - 2018,  International Achievement Scholarship, Portland State University
  • 2010, 2011,  The Second place, Dong Thap Olympiad for High School Student Individual Contest in Math, Vietnam
  • 2010,  Gold Medal, Mekong Delta Olympic in Math, Vietnam
  • 2009,  Silver Medal, Olympic 30/4 in Math, Vietnam

Projects

Homography Estimation
Stack: Python, Pytorch, OpenCV, Numpy, Matplotlib
  • Unofficial implementation of the paper Deep Image Homography Estimation.
  • Estimates a 4-point homography parameterization which maps the four corners from one image into the second image.
  • Achieved average corner error of 6.003 (train) and 6.034 (validation).
Deep Essential Matrix
Stack: Python, Pytorch, OpenCV, Numpy, Matplotlib
  • Estimates the essential matrix between two frames obtained by a monocular camera in epipolar geometry.
  • Uses a ConvNet to directly estimate the matrix without detecting feature points (correlations) between images.
AI Chess
Stack: Python, Numpy, Reinforcement Learning
  • Trained an agent to play chess against human players using reinforcement learning.
  • The agent was trained on playing both sides (black and white) to reduce training time.
  • Result: The agent wins 100/100 games against a Random agent.
AI Minesweeper
Stack: Python, Numpy, Reinforcement Learning
  • Reinforcement learning agent that plays Minesweeper.
  • State representation in the Q-matrix is designed to allow transfer learning from small grids to larger grids.
  • Agent trained on 5x5 grids can successfully play on 9x9 grids.
iOS App
Stack: Swift, K-mean Clustering
  • iOS application that identifies the most dominant color in a photo from the gallery or camera.
  • Utilizes K-mean machine learning algorithm to cluster colors and calculate the average value of the most popular group.
Flappy Bird AI
Stack: Python, NEAT (Neuroevolution)
  • Uses Neuroevolution of Augmenting Topologies (NEAT) to train an agent to play Flappy Bird.
  • The agent achieves perfect play (no losing) after just four generations of evolution.
Imager App
Stack: ReactJS, Django, OpenCV, Heroku
  • Full-stack web application for image processing (Palette generation and Filters).
  • Palette: Detects the n-most dominant colors in an image.
  • Filters: Applies Gaussian blur, detail enhancement, and grayscale effects.