Xingyu Chen (陈星宇)

I am a PhD student at the Inception3D Lab, Westlake University, supervised by Anpei Chen and Andreas Geiger.

I am fortunate to work with Yuliang Xiu at Endless AI Lab and interned at Tencent AI Lab, collaborating with Xuan Wang and Qi Zhang. I got my M.S. from Xi'an Jiaotong University and my B.S. from Chongqing University.

My research topics include computer vision, machine learning, and computer graphics, specifically in spatial intelligence.

Email  |  Google Scholar  |  GitHub  |  Twitter

The world we see is constantly changing: how do intelligent systems generalize to new observations? This question led me to quest for an understanding of the mechanisms underlying spatial intelligence and to develop methods for enabling artificial intelligence with this remarkable capability.

Specifically, I am investigating how generalizability can emerge from reusable 3D & 4D representations, how these representations of the dynamic 3D world could be learned from images & videos, and how inductive biases could serve as expert knowledge to reduce unknown parameters and make learning more efficient.

Equal Contribution *, Corresponding Author †, Project Lead ⚑

A unified framework to probe "texture and geometry awareness" of visual foundation models. Novel view synthesis serves as an effective proxy for 3D evaluation.

We combine local and global alignment via differentiable parameter estimation solvers to achieve robust bundle-adjusting Neural Radiance Fields.

We separate high-frequency human appearance from 3D volume and encode them into a 2D texture, which enables real-time rendering and retexturing.

We decompose the image representation into place, appearance, and occlusion code and use the place code as a descriptor to retrieve images.

Given a sparse semantic map (e.g., pole lines, traffic sign midpoints), we estimate camera poses by learning 2D-3D point-line correspondences.

We recover NeRF from tourism images with variable appearance and occlusions, and consistently render free-occlusion views with hallucinated appearances.

Instead of detecting objects in all frames, we detect only keyframes and embed an efficient prediction mechanism in other frames to find dynamic objects.

We propose a navigation command matching model to discriminate actions generated from sub-optimal policies via smooth rewards.

I am passionate about bridging the physical and digital worlds by building next-generation AR and robotics.

GPS-Denied Navigation, Intelligent Vehicle Future Challenge (IVFC)
Odometry, Mapping, Localization.

Hand gesture recognition
Sensor fusion of IMU and BLE
Robotic hand controller

Teleoperation
Hand gesture recognition
Four-bar linkage structure

Sharing the intuition of dealing with dynamic objects in our previous work and give a prospect of handling the tracking problem via neural fields.

Introduction about Neural Radiance Fields (NeRF) for unconstrained photo collections. Including NeRF, NeRF in the Wild, and Ha-NeRF

• Reviewer of Computer Vision Conferences: CVPR, ICCV, ECCV, 3DV
• Reviewer of Machine Learning Conferences: NeurIPS, ICLR, ICML

template adapted from this awesome website