Ziyi Yang

I am a master student at Zhejiang University advised by professor Xiaogang Jin. My research lies at the neural rendering, inverse rendering, and computer graphics. I am currently interested in 3D Gaussian Splatting (But I am not optimistic about it). Now, I am having a very pleasant time at ByteDance MMLab as a research intern. Prior to joining ZJU, I got my Bachelor's Degree from Shanghai Jiao Tong University in 2022.

I'm now looking for a 25 Fall Ph.D. position in computer graphics / 3D vision! Welcome to contact me!

Email / Google Scholar / Github

Recent News
  • [2024.09] two papers have been accepted by the NeurIPS 2024.
  • [2024.07] one paper has been accepted by the SIGGRAPH Asia 2024 (TOG).
  • [2024.02] two papers have been accepted by the CVPR 2024.
  • [2023.12] one paper has been accepted by the AAAI 2024.
Publications
PontTuset

Spec-Gaussian: Anisotropic View-Dependent Appearance for 3D Gaussian Splatting
Ziyi Yang, Xinyu Gao, Yang-Tian Sun, Yi-Hua Huang, Xiaoyang Lyu, Wen Zhou, Shaohui Jiao, Xiaojuan Qi, Xiaogang Jin
NeurIPS, 2024.
Paper / Project Page / Code

Spec-Gaussian aims to tackle scenes with specular highlights and anisotropy. The key idea is to employ the ASG appearance field instead of SH to model the appearance of 3D Gaussian.

PontTuset

RobIR: Robust Inverse Rendering for High-Illumination Scenes
Ziyi Yang, Yanzhen Chen, Xinyu Gao, Yazhen Yuan, Yu Wu, Xiaowei Zhou, Xiaogang Jin
NeurIPS, 2024.
Project Page / arXiv / code

We propose a novel neural field-based inverse rendering framework for high-illumination scenes. We employ a scene-specific ACES tone mapping and regularized visibility estimation to eliminate the shadow in the PBR materials.

PontTuset

3DGSR: Implicit Surface Reconstruction with 3D Gaussian Splatting
Xiaoyang Lyu, Yang-Tian Sun, Yi-Hua Huang, Xiuzhe Wu, Ziyi Yang, Yilun Chen, Jiangmiao Pang, Xiaojuan Qi
SIGGRAPH Asia (TOG), 2024.
Project Page / arXiv / code

We propose a joint reconstruction technique coupling a GS and neural SDFs to achieve high quality reconstructions.

PontTuset

SC-GS: Sparse-Controlled Gaussian Splatting for Editable Dynamic Scenes
Yi-Hua Huang*, Yang-Tian Sun*, Ziyi Yang*, Xiaoyang Lyu, Yan-Pei Cao, Xiaojuan Qi
CVPR, 2024.
Paper / Project Page / Code

We propose a new representation that explicitly decomposes the motion and appearance of dynamic scenes into sparse control points and dense Gaussians, respectively. Our key idea is to use sparse control points, significantly fewer in number than the Gaussians, to learn compact 6 DoF transformation bases, which can be locally interpolated through learned interpolation weights to yield the motion field of 3D Gaussians. Please visit project page for more demos.

PontTuset

Deformable 3D Gaussians for High-Fidelity Monocular Dynamic Scene Reconstruction
Ziyi Yang, Xinyu Gao, Wen Zhou, Shaohui Jiao, Yuqing Zhang, Xiaogang Jin
CVPR, 2024.
Final score: 5, 5, 5
arXiv / Project Page / Code

The first deformation-based Gaussian splatting for dynamic scenes. We propose a deformable 3D Gaussian Splatting that reconstructs scenes using 3D Gaussians and learns them in canonical space with a deformation field to model monocular dynamic scenes. We also introduce an annealing smoothing training to mitigate the impact of inaccurate poses in real-world datasets.

PontTuset

A General Implicit Framework for Fast NeRF Composition and Rendering
Xinyu Gao, Ziyi Yang, Yunlu Zhao, Yuxiang Sun, Xiaogang Jin, Changqing Zou
AAAI, 2024.
ArXiv

We propose a general implicit pipeline for composing NeRF objects quickly.

Open-source Contribution

1. depth-diff-gaussian-rasterization
Add many extensions to vanilla Gaussian rasterization pipeline used in 3D Gaussian Splatting, including depth forward pass, backward pass, and 4-th SH. Code

2. Awesome-Inverse-Rendering
A collection of papers on NeRF-Based Inverse Rendering. Code

3. floater-free-gaussian-splatting
Fix the densification bug to eliminate floaters in 3D-GS. Code

4. My-exp-Gaussian
Early attempt to model specular highlights with ASG. Code


Last update: 2024.10.09. Thanks.