Chong Bao

I am a Ph.D. student (2020-2026) at the State Key Lab of CAD&CG, Zhejiang University, advised by Prof. Guofeng Zhang and Prof. Zhaopeng Cui. During my Ph.D., I work closely with Yinda Zhang. I used to visit Computer Vision and Geometry Group at ETH Zürich, advised by Marc Pollefeys.

My research interests are in multimodal model, video generation, world model and 3D computer vision.

Email  /  Google Scholar  /  Github

We present a unified multimodal model for holistic digital human generation and understanding across text, audio, motion, semantic video, image, and video.

We scale a DiT-based mixture-of-experts video foundation model with more than 70,000 hours of embodied data for physically grounded video generation and embodied intelligence.

We lift image-editing outputs into continuous, geometry-aware 3D transformations that serve as general priors for open-world robotic manipulation.

We introduce a one-step diffusion model that purifies sparse-view reconstructions corrupted by distractors, improving multi-view consistency and rendering quality.

We introduce a benchmark and multimodal automated evaluation framework for retrieval-augmented visually-rich generation, covering execution correctness, design quality, and content quality.

We propose an Atlanta-world guided, implicit-structured Gaussian representation for efficient and smooth surface reconstruction in indoor and urban scenes.

We present a large-scale synthetic urban dataset with diverse, controllable illumination for benchmarking intrinsic decomposition, inverse rendering, and multi-illumination reconstruction.

We extend neural mesh-based implicit fields with geometry, texture, and semantic controls for versatile, efficient, and interactive volumetric editing.

We propose a new GS with video diffusion model for novel view synthesis and surface reconstruction from extremely sparse(3-4), unposed images in unbounded 360 scenes.

We introduce a novel frequency-aware framework for view synthesis that simultaneously captures the overall scene structure and high-definition details within a single NeRF model.

We present a generic approach to edit 3D avatars in various volumetric representations from a single perspective.

We present a novel neural rendering framework, which is able to learn accurate geometry and reflection of the mirror and support various scene manipulation applications.

We introduce intrinsic decomposition into the NeRF-based neural rendering method and can perform editable novel view synthesis in room-scale scenes.

We present a novel semantic-driven NeRF editing approach, which enables users to edit a neural radiance field with a single image.

We present an efficient covisibility-based incremental SfM and exploit covisibility and registration dependency to describe the image connection

We present a novel mesh-based implicit field with disentangled geometry and texture codes on mesh vertices, which facilitates a set of editing functionalities.

We present a low-cost mapping solution that is able to refine and align the monocular reconstructed point cloud given a public street map.

We present a multi-device integrated cargo loading management system with AR, which monitors cargoes by fusing perceptual information from multiple devices in real-time.

We develop a platform realizing augmented urban reality based on scene localization.

Design and source code from Jon Barron's website