We focus on interdisciplinary work bridging AI, ML, and imaging to solve healthcare, identification, and other real-world challenges.
Developing a robust facial recognition model for player identification across multiple camera feeds in high-speed sports environments, handling motion blur, varying image quality, and long-distance capture.
Developing a multi-modal model for driver identification to ensure verification and enhance security using multimodal sensor data.
Developing a multi-modal model to early detect Alzheimer's in patients using data such as MRIs, cognitive scores, etc.
Developing a novel deep learning model inspired by early development science principles, focusing on constraining and leveraging sample similarity during training to improve neural network performance.
Developing a robust model for the classification and segmentation of fern spores. Identifying distinguishing morphological characteristics.
Investigating representation learning with low-resolution images and efficient training methods for representation learning.
Developing advanced models for early detection of lung nodules using Computed tomography (CT) images.
Developing a lightweight pose estimation model to be used in applications that provide performance analytics for athletes.
Developing a lightweight pose estimation model to detect and analyze unique behavioral patterns in swine following induced mild traumatic brain surgery, aimed at understanding cognitive and motor deficits associated with mild traumatic brain injury.
Developing reinforcement learning techniques to improve training efficiency and performance for non-reinforcement learning models.
Developing an accurate, robust model for classifying the liquid within a surgical drain.
The MARCI Lab will showcase five research posters at the ORNL Core Universities AI Workshop 2025, highlighting work in pose estimation for traumatic brain injury analysis, surgical drain classification, neural network learning periods, multimodal identification frameworks, and SAM2 adaptation for manufacturing applications.
If you would like to reach out, please email: hsantosv@utk.edu
Last updated Oct 15 '25