Research Summary
My group works in the fields of computer vision, machine learning, and deep learning. We usually work in unsupervised or self-supervised settings, with applications in image analysis, video analysis, and time-series analysis. Our recent and current research focuses on:
-
Deep learning models for geometric alignment (for time series, 2D images, or 3D scenes). Example papers:
-
Video analysis. Example papers:
-
Generic, task-agnostic improvements in deep learning, such as wavelet convolutions for large receptive fields and highly-expressive trainable activation functions. Example papers:
-
3D reconstruction and novel view synthesis using Gaussian Splatting (3DGS), typically enhanced by geometric deep features. Example papers:
-
Bayesian nonparametric (BNP) clustering, particularly fast, scalable inference for mixture models and deep models for BNP clustering. Example papers:
-
Fake detection.
-
Underwater computer vision. Example papers:
My earlier work focused on statistics on manifolds (Lie groups and/or Riemannian manifolds) and modeling shape deformations. Example papers:
In even more ancient history, I worked on statistical models for segmentation of MR Brain Images.