Junsheng Zhou | 周俊昇

I am currently a Ph.D. student in School of Software, Tsinghua University, advised by Prof. Yu-Shen Liu.

My research interests lie in the area of 3D computer vision and grapics, especially in 3D reconstruction, generative models, 3D foundation models and cross-modal learning.

Email  /  Google Scholar  /  GitHub

profile photo
News
  • 09/2024: DiffGS, DeepPriorAssembly and Binocular3DGS are accepted to NeurIPS 2024.
  • 06/2024: Our paper FastN2N on fast learning of implicit 3D representations is accepted to TPAMI 2024.
  • 04/2024: The extension of CAP-UDF on implicit representations is accepted to TPAMI 2024.
  • 03/2024: Our paper UDiFF on UDF-based 3D generative models is accepted to CVPR 2024.
  • 03/2024: I will co-organize workshop EMbodied AI: Trends, Challenges, and Opportunities in ICIP 2024.
  • 02/2024: Our paper 3D-OAE on 3D foundation models is accepted to ICRA 2024 for Oral presentation.
  • 01/2024: Our work Uni3D on scaling up 3D foundation models is accepted to ICLR 2024 (Spotlight).
  • 12/2023: Two papers on multi-view reconstruction and point upsampling is accepted to AAAI 2024.
  • 10/2023: Releasing Uni3D, a unified 3D foundation model with one billion parameters.
  • 09/2023: Our paper VP2P-Match on image to LiDAR point cloud registration is accepted to NeurIPS 2023 (Spotlight).
  • 08/2023: Our paper LevelSetUDF on neural implicit representations is accepted to ICCV 2023.
  • 03/2023: Our paper LSA-SDF on implicit representation is accepted to CVPR 2023.
  • 02/2023: Invited to give a talk about implicit surface reconstruction (CAP-UDF) at AI TIME (Video).
  • 11/2022: Our paper NeAF on implicit point normal estimation is accepted to AAAI 2023 (Oral).
  • ---- show more ----
Research

(* Equal Contribution)

DiffGS: Functional Gaussian Splatting Diffusion
Junsheng Zhou*, Weiqi Zhang*, Yu-Shen Liu
Conference on Neural Information Processing Systems (NeurIPS), 2024
project page | arXiv | code

Introducing a powerful 3D generative model that generates Gaussian primitives in arbitrary numbers by functionally disentangling Gaussian Splatting.

Zero-Shot Scene Reconstruction from Single Images with Deep Prior Assembly
Junsheng Zhou, Yu-Shen Liu, Zhizhong Han
Conference on Neural Information Processing Systems (NeurIPS), 2024
project page | arXiv | code

We propose to assemble diverse deep priors from large models for scene generation from single images in a zero shot manner.

Binocular-Guided 3D Gaussian Splatting with View Consistency for Sparse View Synthesis
Liang Han, Junsheng Zhou, Yu-Shen Liu, Zhizhong Han
Conference on Neural Information Processing Systems (NeurIPS), 2024
project page | arXiv | code

We propose a sparse view 3D Gaussian Splatting framework which explores the self supervisions inherent in the binocular stereo consistency.

Fast Learning of Signed Distance Functions from Noisy Point Clouds via Noise to Noise Mapping
Junsheng Zhou*, Baorui Ma*, Yu-Shen Liu, Zhizhong Han
IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), 2024
project page | IEEE Xplore | arXiv | code

We present a fast learning framework capable of inferring signed distance functions from noisy shapes within one minute through noise-to-noise mapping.

UDiFF: Generating Conditional Unsigned Distance Fields with Optimal Wavelet Diffusion
Junsheng Zhou*, Weiqi Zhang*, Baorui Ma, Kanle Shi, Yu-Shen Liu, Zhizhong Han
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024
project page | arXiv | code

UDiFF is a 3D diffusion model for unsigned distance fields (UDFs) which is capable to generate textured 3D shapes with open surfaces from text conditions or unconditionally.

CAP-UDF: Learning Unsigned Distance Functions Progressively from Raw Point Clouds with Consistency-Aware Field Optimization
Junsheng Zhou*, Baorui Ma*, Shujuan Li, Yu-Shen Liu, Yi Fang, Zhizhong Han
IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), 2024
project page | IEEE Xplore | arXiv | code

We present CAP-UDF to represent shapes and scenes with arbitrary architecture by learning a Consistency-Aware unsigned distance function Progressively.

Uni3D: Exploring Unified 3D Representation at Scale
Junsheng Zhou*, Jinsheng Wang*, Baorui Ma*, Yu-Shen Liu, Tiejun Huang, Xinlong Wang
International Conference on Learning Representations (ICLR), 2024 (Spotlight)
Model Zoo | arXiv | code

We present Uni3D, a unified and scalable 3D pretraining framework for large-scale 3D representation learning, and explore its limits at the scale of one billion parameters.

3D-OAE: Occlusion Auto-Encoders for Self-Supervised Learning on Point Clouds
Junsheng Zhou*, Xin Wen*, Baorui Ma, Yu-Shen Liu, Yue Gao, Yi Fang, Zhizhong Han
IEEE International Conference on Robotics and Automation (ICRA), 2024 (Oral)
project page | arXiv | code

We present 3D-OAE, a novel self-supervised point cloud representation learning framework which is highly efficient and can be further transferred to various downstream tasks.

MuSic-UDF: Learning Multi-Scale dynamic grid representation for high-fidelity surface reconstruction from point clouds
Chuan Jin, Tieru Wu, Yu-Shen Liu, Junsheng Zhou
Computers & Graphics (C&G), 2024
project page | arXiv | code

A research project on 3D vision at Jilin University that I served as the advisor.

GeoDream: Disentangling 2D and Geometric Priors for High-Fidelity and Consistent 3D Generation
Baorui Ma*, Haoge Deng*, Junsheng Zhou, Yu-Shen Liu, Tiejun Huang, Xinlong Wang
arXiv, 2024
project page | arXiv | code

We present a 3D generation method that incorporates explicit generalized 3D priors with 2D diffusion priors to enhance the capability of generating unambiguous 3D consistent geometries.

Learning Continuous Implicit Field with Local Distance Indicator for Arbitrary-Scale Point Cloud Upsampling
Shujuan Li*, Junsheng Zhou*, Baorui Ma*, Yu-Shen Liu, Zhizhong Han
AAAI Conference on Artificial Intelligence (AAAI), 2024
project page | arXiv | code

We propose to learn a Local Distance Indicator as local priors to guide arbitrary-scale point cloud upsampling.

NeuSurf: On-Surface Priors for Neural Surface Reconstruction from Sparse Input Views
Han Huang, Yulun Wu, Junsheng Zhou, Ge Gao, Ming Gu, Yu-Shen Liu
AAAI Conference on Artificial Intelligence (AAAI), 2024
project page | arXiv | code

We design a sparse view reconstruction framework that leverages on-surface priors from explicit points to achieve highly faithful surface reconstruction.

Differentiable Registration of Images and LiDAR Point Clouds with VoxelPoint-to-Pixel Matching
Junsheng Zhou*, Baorui Ma*, Wenyuan Zhang, Yi Fang, Yu-Shen Liu, Zhizhong Han
Conference on Neural Information Processing Systems (NeurIPS), 2023 (Spotlight)
project page | arXiv | code

We design a triplet network to learn VoxelPoint-to-Pixel matching via a differentiable probabilistic PnP solver.

Learning a More Continuous Zero Level Set in Unsigned Distance Fields through Level Set Projection
Junsheng Zhou*, Baorui Ma*, Shujuan Li, Yu-Shen Liu, Zhizhong Han
IEEE/CVF International Conference on Computer Vision (ICCV), 2023
project page | arXiv | code

We propose to guide the learning of zero level set in UDF using the rest non-zero level sets via a projection procedure.

Towards Better Gradient Consistency for Neural Signed Distance Functions via Level Set Alignment
Baorui Ma*, Junsheng Zhou*, Yu-Shen Liu, Zhizhong Han
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023
project page | arXiv | code

We propose a level set alignment loss to evaluate the parallelism of level sets, which can be minimized to achieve better gradient consistency and eliminate uncertainty in the field.

Multi-Grid Representation with Field Regularization for Self-Supervised Surface Reconstruction from Point Clouds
Chuan Jin, Tieru Wu, Junsheng Zhou
Shape Modeling International (SMI), 2023     |     Computers & Graphics (C&G), 2023
project page | arXiv | code

A research project on 3D vision at Jilin University that I served as the advisor.

NeAF: Learning Neural Angle Fields for Point Normal Estimation
Shujuan Li*, Junsheng Zhou*, Baorui Ma, Yu-Shen Liu, Zhizhong Han
AAAI Conference on Artificial Intelligence (AAAI), 2023 (Oral)
project page | arXiv | code

We present NeAF, a novel schema to learn implicit angle fields for point normal estimation. NeAF extends the success of implicit functions (e.g. NeRF and DeepSDF) to normal estimation.

Learning Consistency-Aware Unsigned Distance Functions Progressively from Raw Point Clouds
Junsheng Zhou*, Baorui Ma*, Yu-Shen Liu, Yi Fang, Zhizhong Han
Conference on Neural Information Processing Systems (NeurIPS), 2022
project page | arXiv | code

We present CAP-UDF to represent shapes and scenes with arbitrary architecture by learning a Consistency-Aware unsigned distance function Progressively.

3D Shape Reconstruction from 2D Images with Disentangled Attribute Flow
Xin Wen*, Junsheng Zhou*, Yu-Shen Liu, Hua Su, Zhen Dong, Zhizhong Han
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022
project page | arXiv | code

We present 3DAttriFlow to obtain high-quality 2D-to-3D reconstruction and point cloud completion results, while allowing controlled semantic attribute editing.

Honors and Awards
  • National Scholarship (国家奖学金, Top 1% at Tsinghua University), 2023.
  • The 3rd Place in the MVP Completion Challenge (ICCV 2021 Workshop), 2021.
  • Best Poster Paper Awards on ChinaVR, 2021.
Academic Services
  • Co-organizer: "EMbodied AI: Trends, Challenges, and Opportunities" in ICIP-24
  • Program Committee Member: ICLR-24, IJCAI-24, WWW-24, AAAI-25
  • Conference Reviewer: NeurIPS-23/24, ICML-24, CVPR-23/24, ICCV-23, SIGGRAPH Asia-24, ECCV-24, BMVC-23/24, AISTATS-25
  • Journal Reviewer: TOG, TVCG, TIP, CVMJ

template adapted from this awesome website
Last updated: Oct 2024