Fengrui Tian   田豐瑞

I'm a first second-year Ph.D. student at GRASP Lab, University of Pennsylvania, advised by Prof. René Vidal. I have a broad research interest in computer vision and graphics. Now I'm focusing on space generation.

I previously was a research intern at Tsinghua University, where I was very fortunate to work with Prof. Yueqi Duan and focused on novel view synthesis.

I received my M.Eng. degree from the College of Artificial Intelligence of Xi'an Jiaotong University, under the supervision of Prof. Shaoyi Du, with the general ranking 1/107+ and the honor of Best Master's Thesis. I received my B.Eng. degree from the School of Software Engineering of Xi'an Jiaotong University, supervised by Prof. Zhiqiang Tian, with the honor of Best Bachelor's Thesis (1/110+).

Previously, I had some great internships at CCVL group of Johns Hopkins University, Shanghai Artificial Intelligence Laboratory, Sensetime Research and Megvii Research.

I support Slow Science but it is really hard for me to practice it now.

Email  /  Google Scholar  /  Github  /  Full CV (2025.9.16)  /  CV of Failure
profile photo

Selected Publications

DynamicVoyager teaser
Voyaging into Perpetual Dynamic Scenes from a Single View
Fengrui Tian, Tianjiao Ding, Jinqi Luo, Hancheng Min, René Vidal
ICCV, 2025
arxiv  /  paper  /  website  /  code  /  video  /  bibtex
Pose Estimator overview
Learning a Category-level Pose Estimator without Pose Annotations
Fengrui Tian, Yaoyao Liu, Alan Yuille, Adam Kortylewski, Yueqi Duan, Shaoyi Du, Angtian Wang
Arxiv, 2024
arxiv

We propose to leverage the power of diffusion model to train a category-level pose estimator without requiring any pose annotations.

Semantic Flow teaser
Semantic Flow: Learning Semantic Fields of Dynamic Scenes from Monocular Videos
Fengrui Tian, Yueqi Duan, Angtian Wang, Jianfei Guo, Shaoyi Du
ICLR, 2024
open review  /  arxiv  /  code  /  bibtex

We propose Semantic Flow that builds semantic fields of dynamic scenes from monocular videos.

MonoNeRF teaser
MonoNeRF: Learning a Generalizable Dynamic Radiance Field from Monocular Videos
Fengrui Tian, Shaoyi Du, Yueqi Duan
ICCV, 2023
pdf  /  arxiv  /  poster  /  code  /  youtube video  /  talk at CCVL  /  bibtex

We propose MonoNeRF for learning a generalizable dynamic radiance field from monocular videos. While independently using 2D local features and optical flows suffers from ambiguity along the ray direction, they provide complementary constraints to jointly learn 3D point features and scene flows.

TCVM overview
TCVM: Temporal Contrasting Video Montage Framework for Self-supervised Video Representation Learning
Fengrui Tian, Jiawei Fan, Xie Yu, Shaoyi Du, Meina Song, Yu Zhao
ACCV, 2022   (Oral Presentation, Best Paper Award Honorable Mention)
pdf  /  bibtex

We propose a framework for self-supervised video contrastive learning.

Surface-GCN overview
Surface-GCN: Learning Interaction Experience for Organ Segmentation in 3D Medical Images
Fengrui Tian, Zhiqiang Tian, Zhang Chen, Dong Zhang, Shaoyi Du
Medical Physics, 2023   (IF: 4.071)
paper  /  pdf  /  supplementary material  /  bibtex

We propose a framework for learning radiologists' experience in medical organ segmentation.

Academic Services

Conference Reviewer
Journal Reviewer

Honors

Misc

I am a badminton player. I played badminton at UPenn, XJTU, Tsinghua University, SenseTime Corp., and JHU.

Here are several awards I have won:

Don't hesitate to contact me if you want to play badminton with me!


I spent a wonderful time on the Model United Nations activities with my closest friends during my undergraduate study. I appreciate and never forget those memories.

Last update: loading.... The website template was borrowed from the source code and this page.