Shaocheng Yan (颜绍程)

Hey, this is Laka!

🧑‍💻 About Me

Hi, I’m Shaocheng Yan — a first-year Ph.D. student at Wuhan University (M.S. 2023–25 → Ph.D. 2025–), supervised by Prof. Jiayuan Li. My research follows a single question, asked in two steps: given observations, can we reconstruct the world? — and now, can we predict it? From 3D reconstruction to world models, the sensors stay the same; what changes is the ambition.

Outside the lab, you’ll find me on a bicycle 🚴‍♂️ — chasing wind, sunrises 🌅, and sunsets 🌄, wherever the road takes me. When two wheels aren’t enough, I head into the mountains 🏔️ for a good hike 🥾 — no deep thoughts, just legs, trails, and altitude.

Curious to hear from you — research collaborations, half-baked ideas, or trail recommendations.

Currently exploring: 3D Reconstruction · World Models · Robust Geometry · Visual Localization

📖 Scholar 💻 GitHub 📧 Email

🔥 News

2026.04: 🌟 Our paper ULF-Loc is selected as a CVPR'26 ★ Highlight Top ~3%
2026.04: 🎉 Our paper ULF-Loc is accepted by CVPR'26 CCF A
2025.08: 🎙 Invited by 3DCVer to share TurboReg! Replay on
2025.07: 🎙 Invited by CVlife to share TurboReg! Replay on
2025.06: 🎉 Our paper TurboReg is accepted by ICCV'25 CCF A
2025.02: 🎉 Our paper HeMoRa (zh: 赫默拉) is accepted by CVPR'25 CCF A
2024.07: 🎉 Our paper ML-SemReg is accepted by ECCV'24 CCF B

2024.07: 🎉 My homepage is created!

📚 Selected Publications

ULF-Loc: Unbiased Landmark Feature for Robust Visual Localization with 3D Gaussian Splatting

CVPR'26 ★ Highlight Top ~3% CCF A

★ Reviewer scores: 5·5·5·5 (max 6)

$BibTeX$

Yingdong Gu^*, Shaocheng Yan^*, Zhenjun Zhao, Yuan Kou, Jianxin Luo, Pengcheng Shi, Jiayuan Li^†

^* Equal contribution (co-first authors)

TL;DR: Unbiased landmark features via keypoint-consensus sampling and geometry-weighted feature fusion, fundamentally solving alpha-blending bias for highly accurate and ultra-efficient visual localization with 3D Gaussian Splatting.

TurboReg: TurboClique for Robust and Efficient Point Cloud Registration

ICCV'25 CCF A

$BibTeX$

Shaocheng Yan, Pengcheng Shi^†, Zhenjun Zhao, Kaixin Wang, Kuang Cao, Ji Wu, Jiayuan Li^†

TL;DR: A highly efficient and robust estimator for point cloud registration (PCR), supporting both CPU and GPU platforms.

HeMoRa: Unsupervised Heuristic Consensus Sampling for Robust Point Cloud Registration

CVPR'25 CCF A

$BibTeX$

Shaocheng Yan, Yiming Wang, Kaiyan Zhao, Pengcheng Shi, Zhenjun Zhao, Yongjun Zhang, Jiayuan Li^†

TL;DR: Learning a sampling probability distribution for matches in robust estimation — no supervision, reinforcement-inspired.

ML-SemReg: Boosting Point Cloud Registration with Multi-level Semantic Consistency

ECCV'24 CCF B

$BibTeX$

Shaocheng Yan, Pengcheng Shi, Jiayuan Li^†

TL;DR: Boosting 3D keypoint match recall using semantic labels — no learning needed, SLAM-ready.

💻 Selected Projects

TurboReg Star

ICCV'25

TurboClique-based robust & efficient point cloud registration (CPU/GPU).

ULF-Loc Star

CVPR'26

Unbiased landmark features for visual localization with 3D Gaussian Splatting.

ML-SemReg Star

ECCV'24

Multi-level semantic consistency for boosting registration recall.

2025.08

Invited talk on TurboReg @ 3DCVer

2025.07

Invited talk on TurboReg @ CVlife

🏆 Honors and Awards

2025.04, Outstanding Graduate, Class of 2025
2024.11, National Second Prize, China Graduate Mathematical Modeling Contest
2022.08, National First Prize (Champion), China University Robotics Innovation Competition
2020.12, First Prize, 11th National College Student Mathematics Competition (Non-Math Major Category)

🎓 Educations

2025.09 - Present, Ph.D. in Photogrammetry and Remote Sensing, School of Remote Sensing and Information Engineering, Wuhan University
2023.09 - 2025.06, M.S. in Geomatics Engineering, School of Remote Sensing and Information Engineering, Wuhan University
2019.09 - 2023.06, B.S. in Electronic and Information Engineering, School of Electrical Engineering, Southwest Jiaotong University