Siyuan Bian

PhD Student · Computer Vision & Computer Graphics

University of Michigan · Advised by Prof. Jun Gao · Currently interning at NVIDIA

I am a PhD student at the University of Michigan, advised by Prof. Jun Gao, and I am currently interning at NVIDIA. I work at the intersection of computer vision, computer graphics, and digital humans — building models that recover, generate, and edit 3D humans and their clothing from images and language, spanning 3D human pose & shape estimation, garment modeling, and generative approaches that bring large language models into the loop.

My work includes ChatGarment (garment estimation, generation, and editing via LLMs), the HybrIK family of analytical-neural inverse kinematics methods, and NIKI for occlusion-robust human mesh recovery.

News

2025ChatGarment accepted at CVPR 2025 — garment estimation, generation & editing with LLMs.
2025HybrIK-X published in IEEE TPAMI — whole-body mesh recovery.
2024ShapeBoost accepted at AAAI 2024 — part-based human shape estimation.
2023NIKI accepted at CVPR 2023 — occlusion-robust human pose & shape estimation.
2022D&D accepted as an Oral at ECCV 2022.

Publications

* denotes equal contribution. See Google Scholar for the full list.

CVPR 2025

ChatGarment: Garment Estimation, Generation and Editing via Large Language Models

Siyuan Bian, Chenghao Xu, Yuliang Xiu, Artur Grigorev, Zhen Liu, Cewu Lu, Michael J. Black, Yao Feng

A large vision-language model that estimates, generates, and edits sewing-pattern garments directly from images or text instructions.

Paper Project Code Video

IEEE TPAMI 2025

HybrIK-X: Hybrid Analytical-Neural Inverse Kinematics for Whole-body Mesh Recovery

Jiefeng Li, Siyuan Bian, Chao Xu, Zhicun Chen, Lixin Yang, Cewu Lu

Expressive whole-body mesh recovery with articulated hands and face, extending HybrIK's twist-and-swing inverse kinematics.

Paper Project Code

CVPR 2023

NIKI: Neural Inverse Kinematics with Invertible Neural Networks for 3D Human Pose and Shape Estimation

Jiefeng Li*, Siyuan Bian*, Qian Liu, Jiasheng Tang, Fan Wang, Cewu Lu

Invertible networks that model bi-directional errors for occlusion-robust, pixel-aligned human mesh recovery.

Paper Code

AAAI 2024

ShapeBoost: Boosting Human Shape Estimation with Part-Based Parameterization and Clothing-Preserving Augmentation

Siyuan Bian, Jiefeng Li, Jiasheng Tang, Cewu Lu

A part-based shape parameterization and clothing-preserving augmentation for accurate, pixel-aligned body shape recovery.

Paper Code

ECCV 2022 Oral

D&D: Learning Human Dynamics from Dynamic Camera

Jiefeng Li, Siyuan Bian, Chao Xu, Gang Liu, Gang Yu, Cewu Lu

Physics-based 3D human motion reconstruction from in-the-wild video captured by a moving camera.

Paper Project Code

CVPR 2021

HybrIK: A Hybrid Analytical-Neural Inverse Kinematics Solution for 3D Human Pose and Shape Estimation

Jiefeng Li, Chao Xu, Zhicun Chen, Siyuan Bian, Lixin Yang, Cewu Lu

A hybrid inverse-kinematics method that converts accurate 3D keypoints into parametric body meshes via twist-and-swing decomposition.

Paper Project Code

Residual Log-likelihood Estimation teaser

ICCV 2021 Oral

Human Pose Regression with Residual Log-likelihood Estimation

Jiefeng Li, Siyuan Bian, Ailing Zeng, Can Wang, Bo Pang, Wentao Liu, Cewu Lu

A normalizing-flow regression paradigm (RLE) that makes direct pose regression rival heatmap methods, with no test-time overhead.

Paper Code

More publications

arXiv 2025 MetaGen: A DSL, Database, and Benchmark for VLM-Assisted Metamaterial Generation Liane Makatura, Bob Jones, Siyuan Bian, Wojciech Matusik
AAAI 2024 Domain Invariant Learning for Gaussian Processes and Bayesian Exploration Xilong Zhao, Siyuan Bian, Yaoyun Zhang, Yuliang Zhang, Qinya Gu, Xinbing Wang, Chenghu Zhou, Nanyang Ye
ICASSP 2022 Towards Using Clothes Style Transfer for Scenario-Aware Person Video Generation Jingning Xu, Benlai Tang, Mingjie Wang, Siyuan Bian, Wenyi Guo, Xiang Yin, Zejun Ma

Contact

Feel free to reach out about research, collaborations, or questions on any of the work above.