Numair Khan

I am a research scientist at Meta Reality Labs in Redmond, Washington. I work on computer vision and machine learning for applications in computational photography.

I completed my PhD at Brown where I was advised by James Tompkin. I received a Fulbright Scholarship for my Masters at the Courant Institute of New York University where my Master's thesis was advised by Ken Perlin.

Email  /  CV  /  Google Scholar

We propose a method for dynamic scene reconstruction based on a deformable set of 3D Gaussians residing in a canonical space, and a time-dependent deformation field defined by a multi-layer perceptron (MLP)

We present a novel image-guided texture synthesis method to transfer relightable textures from a small number of input images to target 3D shapes across arbitrary categories.

We propose a method for generating tiled multiplane images with only a small number of adaptive depth planes for single-view 3D photography in the wild.

We aim to estimate temporally consistent depth maps of video streams in an online setting by using a global point cloud along with a learned fusion approach in image space.

We present a comprehensive review of neural fields by providing context, mathematical grounding, and an extensive literature review. A companion website contributes a living version that can be continually updated by the community.

A method to estimate dense depth by optimizing a sparse set of points such that their diffusion into a depth map minimizes a multi-view reprojection error from RGB supervision.

We present an algorithm to estimate fast and accurate depth maps from light fields via a sparse set of depth edges and gradients.

We propose a method to compute depth maps for every sub-aperture image in a light field in a view-consistent way.

We use occlusion-aware angular segmentation of an Epipolar Plane Image (EPI) to generate light field superpixels that are consistent across views.

We propose landmark-based verbal directions as an alternative to mini-maps, and examine the development of spatial knowledge in an open-world urban game environment.

Predicting outgoing mobile phone calls using machine learning and time clusters-based approaches.

The distance transform is decomposed into a map-and-reduction pattern for efficient computation on GPUs.

In this measurement study, we analyze whether mobile phone users exhibit temporal regularity in their mobile communication.

The source code for this website was copied from Jon Barron's website. It is freely available for personal use here.