Jeffrey O. Kephart

Jeffrey O. Kephart is affiliated with IBM in the United States and has contributed to research primarily in computer science and engineering. Their work encompasses various subfields including computer vision and pattern recognition, computational mechanics, biomedical engineering, human-computer interaction, and neurology.

Their recent publications are mainly hosted on arXiv (Cornell University) and include the following papers:

• Body Knowledge and Uncertainty Modeling for Monocular 3D Human Body Reconstruction, 2023, arXiv (Cornell University)
• PhysPT: Physics-aware Pretrained Transformer for Estimating Human Dynamics from Monocular Videos, 2024, arXiv (Cornell University)
• A System and Benchmark for LLM-based Q&A on Heterogeneous Data, 2024, arXiv (Cornell University)
• Weakly-Supervised 3D Hand Reconstruction with Knowledge Prior and Uncertainty Guidance, 2024, arXiv (Cornell University)
• Declarative Techniques for NL Queries over Heterogeneous Data, 2025, arXiv (Cornell University)

The main areas of study covered in these works include human pose and action recognition, 3D shape modeling and analysis, hand gesture recognition systems, face recognition and analysis, facial nerve paralysis treatment and research, generative adversarial networks and image synthesis, as well as anatomy and medical technology.

Frequent coauthors in their research collaborations are:

• Qiang Ji
• Zijun Cui
• Yufei Zhang
• Achille Fokoue

Kephart's publication record shows significant focus on advancing techniques in 3D reconstruction, human dynamics estimation, and the integration of knowledge priors and uncertainty modeling in visual computing tasks. Their research also extends into developing systems for natural language queries over diverse and heterogeneous data sources.

Across both main and subfields of study, computer vision plays a central role, alongside applications in biomedical engineering and human-computer interaction. Their contributions span algorithmic innovations and system development aimed at improving the accuracy and efficiency of automated recognition and modeling technologies.

Their works have been disseminated primarily through preprint repositories, facilitating rapid access to evolving research findings. This approach aligns with emerging trends in open scientific communication and interdisciplinary collaboration within the computing and engineering domains.