# Ben Kaye

Source: https://ben-kaye.com/

Computer Vision PhD student at Visual Geometry Group, University of Oxford

I research learned 3D reconstruction, supervised by [Prof. Andrea Vedaldi](https://www.robots.ox.ac.uk/~vedaldi/) and [Prof. Christian Rupprecht](https://chrirupp.github.io/).

I’m looking for a research internship for summer 2027.

## Links

- [Email](mailto:ben@ben-kaye.com)
- [GitHub](https://github.com/ben-kaye)
- [Scholar](https://scholar.google.com/citations?user=2alhGi4AAAAJ)
- [CV](https://ben-kaye.com/cv)

- [Breathing Timer](https://ben-kaye.com/breathe/)

Location: Oxford, UK

## Publications

### DualPM: Dual Posed-Canonical Point Maps for 3D Shape and Pose Reconstruction

CVPR 2025 · Highlight

Ben Kaye (equal contribution), [Tomas Jakab](https://www.robots.ox.ac.uk/~tomj/) (equal contribution), [Shangzhe Wu](https://www.elliottwu.com/), [Christian Rupprecht](https://chrirupp.github.io/), [Andrea Vedaldi](https://www.robots.ox.ac.uk/~vedaldi/)

Single-image 3D shape and pose recovery for deformable objects such as animals. By predicting paired posed and canonical point maps, deformation falls out as a difference. We train on synthetic data and get state-of-the-art results on real images.

- [Project page](https://dualpm.github.io/)
- [arXiv](https://arxiv.org/abs/2412.04464)
- [Code](https://github.com/DualPM/DualPM_Paper)