Dongbo Shi
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Dongbo Shi has a total of 13 co-authors. Here are the top co-authors by number of shared publications:
- Renjie Chen
School of Mathematical Sciences, University of Science and Technology of China, Hefei, China
3 shared publications - Bojian Wu
Tencent Games
2 shared publications - Jieping Ye
Qwen-Character Team, Alibaba Group, China
2 shared publications - Jinhui Guo
Independent
2 shared publications - Ligang Liu
University of Science and Technology of China, Hefei, China
2 shared publications - Lubin Fan
Alibaba Cloud Computing Yuyang Hong
2 shared publications
Latest Publications
- Open Access
- 27 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in TechRxiv
A Survey on Pose-Free Neural Radiance Fields and 3D Gaussian Splatting
Neural Radiance Fields (NeRF) and 3D Gaussian Splatting (3DGS) have significantly advanced 3D scene reconstruction and novel view synthesis in recent years. Despite continuous improvements in accuracy, quality, and application scope, these methods fundamentally rely on a critical assumption: precise camera poses for all input images. This requirement substantially diverges from practical scenarios, where factors such as motion blur from camera movement, object occlusion in complex scenes, visual changes in dynamic environments, pose estimation difficulties in large-scale scenes, and hardware limitations of acquisition devices can hinder accurate camera pose estimation. This survey presents the first comprehensive review of pose-free NeRF and 3DGS research, tackling this fundamental challenge. We survey the extensive literature, conduct an in-depth analysis of methodological innovations, and categorize advances into three key directions for systematic discussion: 1) base model enhancement, 2) training strategy refinement, and 3) novel prior incorporation. We further collect benchmark resources (including datasets and metrics) for evaluating pose-free methods and analyze the performance of representative works. Finally, we discuss core challenges and emerging solutions for real-world applications, highlighting promising avenues for future research to advance the field.
- 27 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in TechRxiv
Open AccessA Survey on Pose-Free Neural Radiance Fields and 3D Gaussian Splatting
Neural Radiance Fields (NeRF) and 3D Gaussian Splatting (3DGS) have significantly advanced 3D scene reconstruction and novel view synthesis in recent years. Despite continuous improvements in accuracy, quality, and application scope, these methods fundamentally rely on a critical assumption: precise camera poses for all input images. This requirement substantially diverges from practical scenarios, where factors such as motion blur from camera movement, object occlusion in complex scenes, visual changes in dynamic environments, pose estimation difficulties in large-scale scenes, and hardware limitations of acquisition devices can hinder accurate camera pose estimation. This survey presents the first comprehensive review of pose-free NeRF and 3DGS research, tackling this fundamental challenge. We survey the extensive literature, conduct an in-depth analysis of methodological innovations, and categorize advances into three key directions for systematic discussion: 1) base model enhancement, 2) training strategy refinement, and 3) novel prior incorporation. We further collect benchmark resources (including datasets and metrics) for evaluating pose-free methods and analyze the performance of representative works. Finally, we discuss core challenges and emerging solutions for real-world applications, highlighting promising avenues for future research to advance the field.
- Open Access
NoPe-NeRF++: Local-to-Global Optimization of NeRF with No Pose Prior
In this paper, we introduce NoPe-NeRF++, a novel local-to-global optimization algorithm for training Neural Radiance Fields (NeRF) without requiring pose priors. Existing methods, particularly NoPe-NeRF, which focus solely on the local relationships within images, often struggle to recover accurate camera poses in complex scenarios. To overcome the challenges, our approach begins with a relative pose initialization with explicit feature matching, followed by a local joint optimization to enhance the pose estimation for training a more robust NeRF representation. This method significantly improves the quality of initial poses. Additionally, we introduce global optimization phase that incorporates geometric consistency constraints through bundle adjustment, which integrates feature trajectories to further refine poses and collectively boost the quality of NeRF. Notably, our method is the first work that seamlessly combines the local and global cues with NeRF, and outperforms state-of-the-art methods in both pose estimation accuracy and novel view synthesis. Extensive evaluations on benchmark datasets demonstrate our superior performance and robustness, even in challenging scenes, thus validating our design choices.
Open AccessNoPe-NeRF++: Local-to-Global Optimization of NeRF with No Pose Prior
In this paper, we introduce NoPe-NeRF++, a novel local-to-global optimization algorithm for training Neural Radiance Fields (NeRF) without requiring pose priors. Existing methods, particularly NoPe-NeRF, which focus solely on the local relationships within images, often struggle to recover accurate camera poses in complex scenarios. To overcome the challenges, our approach begins with a relative pose initialization with explicit feature matching, followed by a local joint optimization to enhance the pose estimation for training a more robust NeRF representation. This method significantly improves the quality of initial poses. Additionally, we introduce global optimization phase that incorporates geometric consistency constraints through bundle adjustment, which integrates feature trajectories to further refine poses and collectively boost the quality of NeRF. Notably, our method is the first work that seamlessly combines the local and global cues with NeRF, and outperforms state-of-the-art methods in both pose estimation accuracy and novel view synthesis. Extensive evaluations on benchmark datasets demonstrate our superior performance and robustness, even in challenging scenes, thus validating our design choices.
- 12 April 2025
- Published by Wiley in Computer Graphics Forum
- Vol. 44 (2)
NoPe‐NeRF++: Local‐to‐Global Optimization of NeRF with No Pose Prior
In this paper, we introduce NoPe‐NeRF++, a novel local‐to‐global optimization algorithm for training Neural Radiance Fields (NeRF) without requiring pose priors. Existing methods, particularly NoPe‐NeRF, which focus solely on the local relationships within images, often struggle to recover accurate camera poses in complex scenarios. To overcome the challenges, our approach begins with a relative pose initialization with explicit feature matching, followed by a local joint optimization to enhance the pose estimation for training a more robust NeRF representation. This method significantly improves the quality of initial poses. Additionally, we introduce global optimization phase that incorporates geometric consistency constraints through bundle adjustment, which integrates feature trajectories to further refine poses and collectively boost the quality of NeRF. Notably, our method is the first work that seamlessly combines the local and global cues with NeRF, and outperforms state‐of‐the‐art methods in both pose estimation accuracy and novel view synthesis. Extensive evaluations on benchmark datasets demonstrate our superior performance and robustness, even in challenging scenes, thus validating our design choices.
- 12 April 2025
- Published by Wiley in Computer Graphics Forum
- Vol. 44 (2)
NoPe‐NeRF++: Local‐to‐Global Optimization of NeRF with No Pose Prior
In this paper, we introduce NoPe‐NeRF++, a novel local‐to‐global optimization algorithm for training Neural Radiance Fields (NeRF) without requiring pose priors. Existing methods, particularly NoPe‐NeRF, which focus solely on the local relationships within images, often struggle to recover accurate camera poses in complex scenarios. To overcome the challenges, our approach begins with a relative pose initialization with explicit feature matching, followed by a local joint optimization to enhance the pose estimation for training a more robust NeRF representation. This method significantly improves the quality of initial poses. Additionally, we introduce global optimization phase that incorporates geometric consistency constraints through bundle adjustment, which integrates feature trajectories to further refine poses and collectively boost the quality of NeRF. Notably, our method is the first work that seamlessly combines the local and global cues with NeRF, and outperforms state‐of‐the‐art methods in both pose estimation accuracy and novel view synthesis. Extensive evaluations on benchmark datasets demonstrate our superior performance and robustness, even in challenging scenes, thus validating our design choices.
- 1 August 2022
- Published by Wiley in Computer Graphics Forum
- Vol. 41 (5) , 1-11
Harmonic Shape Interpolation on Multiply‐connected Planar Domains
- 1 August 2022
- Published by Wiley in Computer Graphics Forum
- Vol. 41 (5) , 1-11
Harmonic Shape Interpolation on Multiply‐connected Planar Domains