Wesley Khademi

I am a second-year Ph.D. student in Computer Science at Oregon State University, advised by Fuxin Li.

I received my bachelor's degree in Computer Science with a minor in Mathematics from California Polytechnic State University, San Luis Obispo. At Cal Poly, I worked on problems related to self-supervised denoising and novel view synthesis with Jonathan Ventura and neural networks for solving PDEs with Stathis Charalampidis.

Email  /  CV  /  Google Scholar  /  Github

profile photo

My research interests are in the fields of computer vision and machine learning. In particular, I am interested in how various visual modalities (i.e, RGB images, point clouds) can be leveraged to reason about and recreate our 3D world. Most recently, my work has focused on generative modeling of point clouds for 3D shape generation and completion.

I am always interested in meeting and collaborating with new people who share common research interests as me. Please feel free to reach out to me via email to chat about previous or future work.


[09/2023] Our paper "Diverse Shape Completion via Style Modulated Generative Adversarial Networks" was accepted at NeurIPS 2023.

[06/2023] Joined Reality Labs Research, Meta, as a Research Scientist Intern for Summer & Fall 2023.

[06/2023] Successfully passed my Ph.D. qualifying exam.

[04/2021] Accepted a Ph.D. offer at Oregon State University.

Neural Networks Enforcing Physical Symmetries in Nonlinear Dynamical Lattices: The Case Example of the Ablowitz-Ladik Model
Wei Zhu, Wesley Khademi, Efstathios G. Charalampidis, Panayotis G. Kevrekidis.
Physica D: Nonlinear Phenomena, 2022
arXiv / bibtex

We introduce symmetry-preserving, physics-informed neural networks (S-PINNs) motivated by symmetries that are ubiquitous to solutions of nonlinear dynamical lattices. Through the correlation of parity symmetries in both space and time of solutions to differential equations with their group equivariant representation, we construct group-equivariant NNs which respect spatio-temporal parity symmetry. Moreover, we adapt the proposed architecture to enforce different types of periodicity (or localization) of solutions to nonlinear dynamical lattices.

View Synthesis In Casually Captured Scenes Using a Cylindrical Neural Radiance Field With Exposure Compensation
Wesley Khademi, Jonathan Ventura.
ACM SIGGRAPH Posters, 2021
project page / abstract / poster / bibtex

We extend Neural Radiance Fields (NeRF) with a cylindrical parameterization and learned exposure compensation technique that enables rendering photorealistic novel views of casually captured 360-degree outward facing scenes.

Self-Supervised Poisson-Gaussian Denoising
Wesley Khademi, Sonia Rao, Clare Minnerath, Guy Hagen, Jonathan Ventura.
arXiv / code / bibtex

We extend the blindspot model for self-supervised denoising to handle Poisson-Gaussian noise and introduce an improved training scheme that avoids hyperparameters and adapts the denoiser to the test data.

Thank you to Jon Barron for the website template.