Jim T. James

Hi! I’m Jim, a PhD student in Machine Learning at Georgia Tech advised by James Hays. My research interests lie at the intersection of computer vision and machine learning. My current research involves developing parameter efficient encoders for 3D perception.

Prior to the start of my PhD, I received my Bachelor’s degree in Computer Science from Georgia Tech, with a concentration in artificial intelligence and theory.

Outside of academica, I’m an avid amateur astronomer. I enjoy both visual observation from a dobsonian and astrophotography with camera lenses. Most recently, I’ve begun an interest in designing my own telescopes with hand-ground optics and precision mounts. During the daytime, I turn my lenses towards the birds.

selected publications

UAI 2024

One Shot Inverse Reinforcement Learning for Stochastic Linear Bandits

Etash Kumar Guha, Jim James, Krishna Acharya, and 2 more authors

In The 40th Conference on Uncertainty in Artificial Intelligence, 2024
Comput. Mater. Sci.

Segmentation of tomography datasets using 3D convolutional neural networks

Jim James, Nathan Pruyne, Tiberiu Stan, and 6 more authors

Computational Materials Science, 2023

Abs

Dendritic microstructures are ubiquitous in nature and are the primary solidification morphologies in metallic materials. Techniques such as X-ray computed tomography (XCT) have provided new insights into dendritic phase transformation phenomena. However, manual identification of dendritic morphologies in microscopy data can be both labor intensive and potentially ambiguous. The analysis of 3D datasets is particularly challenging due to their large sizes (terabytes) and the presence of artifacts scattered within the imaged volumes. In this study, we trained 3D convolutional neural networks (CNNs) to segment 3D datasets. Three CNN architectures were investigated, including a new version of FCDenseNet which we extended to 3D. We show that using hyperparameter optimization (HPO) and fine-tuning techniques, both 2D and 3D CNN architectures outperform the previous state of the art. The 3D U-Net architecture trained in this study produced the best segmentations according to quantitative metrics (intersection-over-union of 95.56% and a boundary displacement error of 0.58 pixels), while 3D FCDense produced the smoothest boundaries and best segmentations according to visual inspection. The trained 3D CNNs are able to segment entire 852 × 852 × 250 voxel 3D volumes in only ∼60 s, thus hastening the progress towards a deeper understanding of phase transformation phenomena such as dendritic solidification.