About Me
I am a Ph.D. candidate in Computer Science at the University of Maryland, specializing in 3D computer vision, point cloud understanding, and generative AI. My research aims to develop advanced techniques for interpreting and generating 3D data, with applications in robotics, autonomous systems, and digital twins. I am particularly interested in leveraging generative models like diffusion models and autoencoders to enhance 3D place recognition and scene understanding. As a contributor to projects like CROSSLOC3D, I have gained expertise in cross-source data fusion and deep learning for 3D vision. My work at NASA Goddard and ARLIS has further strengthened my skills in scientific programming and AI applications.
News
- April 2025 – Preparing for NeurIPS 2025 submission
- October 2023 – Published CROSSLOC3D at ICCV 2023
Research Interests
- 3D Computer Vision
- LiDAR and Point Clouds
- Generative Models (Diffusion, Autoencoders)
- Robotics
- Deep Learning
Education
- Ph.D., Computer Science — UMD (2022–2026, GPA: 3.9)
- M.S., Computer Science — UMD (2021–2022, GPA: 3.9)
- B.S., CS & Math — UMD (2018–2021, GPA: 3.86)
- A.S., Computer Science — AACC (2015–2018, GPA: 3.9)
Research Experience
- Artificial Intelligence Researcher, Applied Research Laboratory for Intelligence and Security (ARLIS), UMD (Jul 2022 – Present)
- Designing and developing 3D spatial software tools for digital twin applications in defense and intelligence.
- Previously applied reinforcement learning to optimize supply chain simulations using AnyLogic.
- Graduate Researcher, University of Maryland, College Park (2021 – Present)
- Contributor to CROSSLOC3D (ICCV 2023), a cross-source 3D place recognition framework.
- Research focus on generative modeling with latent diffusion and autoencoder architectures for 3D data.
- Lead developer of the Sand Browser and Spatial Data Structure Visualizations (Spatial Index Demos: https://donar.umiacs.umd.edu/quadtree/).
- Contributed to research on adversarial attacks against MAML in few-shot learning contexts.
- Developed a differentiable agent-based traffic simulation using JAX for accelerated computation.
Professional Experience
- Mass Spectrometer Calibration Intern, NASA Goddard Space Flight Center, Greenbelt, MD (Mar 2021 – Jun 2021)
- Developed C++ code for data acquisition and calibration of mass spectrometers.
- Worked with DAQs, Raspberry Pi, and ADCs for sensor alignment and signal processing.
- Supported instrument software for ion spectrometry and real-time telemetry.
Publications
Projects
- CROSSLOC3D: Developed a diffusion-based 3D place recognition system that achieves state-of-the-art performance in cross-source aerial-ground matching. [Code]
- Adversarial Attacks on MAML: Investigated transferable patch attacks and watermark poisoning techniques in few-shot learning settings to assess the robustness of meta-learning algorithms.
- JAX-based Traffic Simulation: Integrated JAX into a traffic simulation framework to enable GPU acceleration and gradient-based optimization, significantly improving simulation speed and enabling new research directions.
- Secure ATM System: Designed and implemented a secure ATM system in C, incorporating OpenSSL for encryption and security modeling to prevent man-in-the-middle attacks.
Skills
Python, C++, Java, R, SQL, PyTorch, TensorFlow, JAX, Flask, Docker, Linux
Honors & Affiliations
- Mortar Board Senior Honor Society (Adele H. Stamp Chapter)
- Phi Theta Kappa Honor Society (Omicron Theta Chapter)
- Dean’s List (All semesters since 2018)
Contact
Location: Annapolis, MD 21409
Email: mjhoover@umd.edu
Phone: (443) 852-8655
GitHub: github.com/mjhoover1
LinkedIn: linkedin.com/in/montana-hoover