Junsheng Zhou | 周俊昇

I am currently a Master student (since fall, 2021) and an incoming 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, 3D foundation models, cross-modal learning and generative models.

Email  /  Google Scholar  /  Github

• 01/2024: Our work Uni3D on scaling up 3D foundation models got accepted to ICLR 2024 (Spotlight).
• 12/2023: Two papers on multi-view reconstruction and point upsampling got 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 got accepted to NeurIPS 2023 (Spotlight).
• 08/2023: Our paper LevelSetUDF on neural implicit representations got accepted to ICCV 2023.
• 06/2023: One paper on surface reconstruction that I advised got accepted to SMI 2023 and C&G 2023.
• 03/2023: Our paper LSA-SDF on implicit representation got 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 got accepted to AAAI 2023 (Oral).
• 09/2022: Our paper CAP-UDF on surface reconstruction got accepted to NeurIPS 2022.
• 03/2022: Our paper 3DAttriFlow on point cloud generation got accepted to CVPR 2022.
• 11/2021: My team won the 3rd Place in the MVP Completion Challenge (ICCV 2021 Workshop).
• 10/2021: Our paper MSM-Vis on data visualization won the Best Poster Paper on ChinaVR 2021.
• 09/2021: Started master journey at Tsinghua University.

(* Equal Contribution)

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.

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.

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

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

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

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

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.

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

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.

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

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

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.

• 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.

• Program Committee Member: ICLR-24, IJCAI-24, WWW-24
• Conference Reviewer: NeurIPS-23, CVPR-23/24, ICCV-23, ECCV-24, BMVC-23
• Journal Reviewer: CVMJ

template adapted from this awesome website
Last updated: Jan 2024