Wenzheng Chen

I'm a tenure-track Assistant Professor at Wangxuan Institute of Computer Technology, Peking University. My research focuses on computational photography, 3D vision, and spatial intelligence.

Before joining Peking University, I was a Research Scientist at NVIDIA Toronto AI Lab. I earned my Ph.D. from the University of Toronto and received both my Master's and Bachelor's degrees from Shandong University.

I am currently recruiting Ph.D. students and research interns. Please see the Hiring section below for details.

Email  /  CV  /  Google Scholar  /  LinkedIn

Full publication list in Google Scholar.

Neural single-shot Structured Light (NSL) achieves robust and high-fidelity 3D reconstruction from a single-shot structured light input.

GeoSplatting introduces a novel hybrid representation that grounds 3DGS with isosurfacing to provide accurate geometry and normals for high-fidelity inverse rendering.

RainyGS integrates physics simulation with 3DGS to efficiently generate photorealistic, physically accurate, and controllable dynamic rain effects for in-the-wild scenes.

EMDiffusion learns a clean diffusion model from corrupted data.

TurboSL provides sub-pixel-accurate surfaces and normals at mega-pixel resolution from structured light images, captured at fractions of a second.

Combined with other NVIDIA technology, FEGR is one component of Neural Reconstruction Engine announced in GTC Sept 2022 Keynote.

We develop a 3D generative model to generate meshes with textures, bridging the success in the differentiable surface modeling, differentiable rendering and 2D GANs.

We propose a hybrid lighting representation to represent spatial-varying lighting for complex outdoor street scenes.

Nvdiffrec reconstructs 3D mesh with materials from multi-view images by combining diff surface modeling with diff renderer. The method supports Nvidia Neural Drivesim

DIB-R++ is a highly performant differentiable renderer that combines rasterization and ray tracing and supports advanced lighting and material effects. We further embed it in deep learning and jointly predict geometry, texture, lighting, and material from a single image.

We explore StyleGAN as a multi-view image generator and train inverse graphics from StyleGAN images. Once trained, the inverse graphics model further helps disentangle and manipulate StyleGAN latent code from graphics knowledge. Our work was featured at NVIDIA GTC 2021 and has become an Omniverse product.

We propose to learn feature embeddings for non-line-of-sight imaging and recognition by propagating features through physical modules.

We predict deformable tetrahedral meshes from images or point clouds, which support arbitrary topologies. We also design a differentiable renderer for tetrahedra, allowing 3D reconstruction from 2D supervision only.

We present optical SGD, a computational imaging technique that allows an active depth imaging system to automatically discover optimal illuminations & decoding.

An interpolation-based differentiable 3D mesh renderer that supports vertex positions, vertex colors, multiple lighting models, and texture mapping, and can be easily embedded in neural networks.

We show that hidden objects can be recovered from conventional images instead of transient images.

We predict object polygon contours from graph neural networks, where a novel 2D differentiable rendering loss is introduced. It renders a polygon contour into a segmentation mask and backpropagates the loss to help optimize the polygon vertices.

À La Carte designs structured-light patterns from a machine-learning perspective, where patterns are automatically optimized by minimizing the disparity error under any given imaging condition.

3D pose estimation using a model trained with synthetic data and domain adaptation.