Geo4D: Leveraging Video Generators for Geometric 4D Scene Reconstruction
ICCV, 2025
A method to repurpose video diffusion models for monocular 3D reconstruction of dynamic scenes.
I am currently a Marie Skłodowska-Curie Actions (MSCA) Fellow in VGG at the University of Oxford, working with Andrea Vedaldi on feed-forward photorealistic 3D and 4D reconstruction. I was very fortunate to also collaborate with Andrew Zisserman, Christian Rupprecht, and Iro Laina in VGG. I received my Ph.D. degree in Computer Science from the Nanyang Technological University (NTU), advised by Prof. Tat-Jen Cham and Prof. Jianfei Cai.
I am an incoming Nanyang Assistant Professor (starting from Fall 2025) at the College of Computing and Data Science, Nanyang Technological University, where I lead the Physical Visual Group. My research focuses on Creative AI, aiming to develop systems that perceive, reconstruct and interact with the physical world. The broader goal is to create realistic digital twins of the natural world, with various physical properties, capturing not only appearance, content, and geometry but also occlusion, dynamics, gravity, interaction, sound and more.
📍Openings: more details are available at Position.
- Phd student: we have multiple PhD openings for Spring 2026 and Fall 2026.
- Postdoc position: we have two postdoc positions available.
- Research Assistant: we have two research assistant positions available.
Selected Publications
More Publications and Google Scholar
2025
arxivDSO: Aligning 3D Generators with Simulation Feedback for Physical Soundness
ICCV, 2025
A feed-forward approach to increase the likelihood that the 3D generator outputs stable 3D objects directly.
Amodal3R: Amodal 3D Reconstruction from Occluded 2D Images
ICCV, 2025
A modal for completed 3D reconstruction from partially visible inputs.
Puppet-Master: Scaling Interactive Video Generation as a Motion Prior for Part-Level Dynamics
ICCV, 2025
An interactive video generative model that can serve as a motion prior for part-level dynamics.
Flash3D: Feed-Forward Generalisable 3D Scene Reconstruction from a Single Image
3DV, 2025
Stanislaw Szymanowicz*, Eldar Insafutdinov*, Chuanxia Zheng*, Dylan Campbell, João Henriques, Christian Rupprecht, Andrea Vedaldi
A fast, super efficient, trainable on a single GPU in one day for scene 3D reconstruction from a single image.
2024
A General Protocol to Probe Large Vision Models for 3D Physical Understanding
NeurIPS, 2024
Probe large vision models to determine to what extent they 'understand' different physical properties in an image.
MVSplat360: Feed-Forward 360 Scene Synthesis from Sparse Views
NeurIPS, 2024
A feed-forward approach for 360-degree scene-level novel view synthesis using only sparse observations.
MVSplat: Efficient 3D Gaussian Splatting from Sparse Multi-View Images
ECCV, 2024(oral)
Yuedong Chen, Haofei Xu, Chuanxia Zheng, Bohan Zhuang, Marc Pollefeys, Andreas Geiger, Tat-Jen Cham, Jianfei Cai
A feed-forward approach for efficiently predicting 3D Gaussians from sparse multi-view images in a single forward pass.
DragAPart: Learning a Part-Level Motion Prior for Articulated Objects
ECCV, 2024
A physical interaction with objects in vision for part-level dragging.
Free3D: Consistent Novel View Synthesis without 3D Representation
CVPR, 2024
A method to synthesize consistent novel views from a single image on open-set categories without the need of explicit 3D representations.
PanoDiffusion: 360-degree Panorama Outpainting via Diffusion
ICLR, 2024
An indoor panorama outpainting model using latent diffusion models with view-consistent.
2023
Online clustered codebook
ICCV, 2023
A simple approach to avoid codebook collapse and achive 100% codebook utilisation.
UniD3: Unified Discrete Diffusion for Simultaneous Vision-Language Generation
ICLR, 2023
Minghui Hu, Chuanxia Zheng, Heliang Zheng, Tat-Jen Cham, Chaoyue Wang, Zuopeng Yang, Dacheng Tao, P.N.Suganthan
A unified discrete diffusion model for simultaneous vision-language generation.
2022
MoVQ: Modulating Quantized Vectors for High-Fidelity Image Generation
NeurIPS, 2022(spotlight)
A spatially conditional normalization is introduced to address the repeated artifacts in vector quantized methods.
Sem2NeRF: Converting Single-View Semantic Masks to Neural Radiance Fields
ECCV, 2022
A model that transfers the 2D semantic map into 3D NeRF, and lets users edit 3D model through 2D semantic input.
Bridging global context interactions for high-fidelity image completion
CVPR, 2022
TFill fills in reasonable contents for both foreground object removal and content completion.
2021
Visiting the Invisible: Layer-by-Layer Completed Scene Decomposition
IJCV, 2021
A high-level scene understanding system that simultaneously models the completed shape and appearance for all instances.
The Spatially-Correlative Loss for Various Image Translation Tasks
CVPR, 2021
A novel spatially-correlative loss that is simple, efficient and yet effective for preserving scene structure consistency while supporting large appearance changes during unpaired I2I translation.
2019
Pluralistic (Free-Form) Image Completion
CVPR, 2019IJCV, 2021
Given a masked image, the proposed pic model is able to generate multiple and diverse plausible results.
2018
T2Net: Synthetic-to-Realistic Translation for Depth Estimation Tasks
ECCV, 2018
Without using any real depth map, the proposed model evaluates depth maps on real scenes using only synthetic datasets.
Professional Services
- Area chair: BMVC (2024-2025), ACM MM 2024
- Conference reviewer: CVPR (2020-2025), ICCV (2019-2025), ECCV (2020-2024), NeurIPS (2022-2025), ICLR (2021-2025), ICML 2023
- Journal reviewer: TPAMI, IJCV, TIP, PR, TMM (Outstanding Reviewer Award, 2021), TCSVT
Awards and Honors
- Singapore NRF Fellow, National Research Foundation, 2025
- DAAD Ainet Fellow, 2024
- Marie Skłodowska-Curie Actions (MSCA) Fellow, 2024
- NTU Outstanding PhD Thesis Award, 2022