Xinze Li

Undergraduate Student Beijing Normal-Hong Kong Baptist University Email: t330026083@mail.bnbu.edu.cn

I am currently an undergraduate student at Beijing Normal-Hong Kong Baptist University (BNBU) and am looking for a PhD position for Fall 2027.

Before joining BNBU in September 2023, I received systematic training in Olympiad in Informatics at Changjun High School beginning in 2017.

My previous research interests were in bioinformatics and medical image analysis, where I completed four research projects as both an independent and collaborative researcher. I also worked as a research assistant with Central South University.

Since Spring 2025, I have focused on 3D Computer Vision under the supervision of Dr. Wentao Cheng, with a particular focus on:

3D Reconstruction
Efficient Feed-Forward 3D Foundation Models

news

Aug 18, 2026	I joined the Autonomous Driving department at Horizon Robotics as a Research Intern.
Jul 02, 2026	I was honored to receive the Outstanding Reviewer Award for ICME 2026.
Jun 24, 2026	We released the Project Page for ReCal3R: a reliability-calibrated learning-rate rule for streaming 3D reconstruction.
May 01, 2026	One paper titled “QuadBox: Accelerating 3D Gaussian Splatting with Geometry-Aware Boxes” was accepted to ICIP 2026.
Mar 18, 2026	Two papers on 3D Foundation Models and Local Feature Learning have been accepted by ICME 2026! 🎉🎉 S-VGGT: Structure-Aware Subscene Decomposition for Scalable 3D Foundation Models. D2Feat: Dual Distillation for Semantic and Geometric Local Feature Learning. Stay tuned for the full papers!

selected publications

ICIP

QuadBox: Accelerating 3D Gaussian Splatting with Geometry-Aware Boxes

Xinze Li, Bohan Yang, Pengxu Chen, and 4 more authors

In 2026 IEEE International Conference on Image Processing (ICIP), 2026

DOI Bib PDF Code

@inproceedings{li2026quadbox,
  title = {QuadBox: Accelerating 3D Gaussian Splatting with Geometry-Aware Boxes},
  author = {Li, Xinze and Yang, Bohan and Chen, Pengxu and Wang, Yiyuan and Luo, Hongcheng and Cheng, Wentao and Su, Weifeng},
  booktitle = {2026 IEEE International Conference on Image Processing (ICIP)},
  year = {2026},
  pages = {1-6},
  doi = {10.1109/ICIP61757.2026.11629922},
}

arXiv

S-VGGT: Structure-Aware Subscene Decomposition for Scalable 3D Foundation Models

Xinze Li, Pengxu Chen, Yiyuan Wang, and 2 more authors

arXiv preprint arXiv:2603.17625, 2026

Bib PDF Code

@article{li2026svggt,
  title = {S-VGGT: Structure-Aware Subscene Decomposition for Scalable 3D Foundation Models},
  author = {Li, Xinze and Chen, Pengxu and Wang, Yiyuan and Su, Weifeng and Cheng, Wentao},
  journal = {arXiv preprint arXiv:2603.17625},
  year = {2026},
}

ICME

D2Feat: Dual Distillation for Semantic and Geometric Local Feature Learning

Yiyuan Wang, Xinze Li, Puzhen Wu, and 3 more authors

In IEEE International Conference on Multimedia and Expo (ICME), 2026

Bib

@inproceedings{wang2026d2feat,
  title = {D2Feat: Dual Distillation for Semantic and Geometric Local Feature Learning},
  author = {Wang, Yiyuan and Li, Xinze and Wu, Puzhen and Zhang, Junkai and Su, Weifeng and Cheng, Wentao},
  booktitle = {IEEE International Conference on Multimedia and Expo (ICME)},
  year = {2026},
}

arXiv

ReCal3R: Reliability-Calibrated Learning Rates for Streaming 3D Reconstruction

Xinze Li, Yiyuan Wang, Pengxu Chen, and 3 more authors

arXiv preprint arXiv:2607.05356, 2026

Bib PDF Code

@article{li2026recal3r,
  title = {ReCal3R: Reliability-Calibrated Learning Rates for Streaming 3D Reconstruction},
  author = {Li, Xinze and Wang, Yiyuan and Chen, Pengxu and Su, Weifeng and Lin, Weisi and Cheng, Wentao},
  journal = {arXiv preprint arXiv:2607.05356},
  year = {2026},
}