NGS-Marker: Robust Native Watermarking for 3D Gaussian Splatting
Robust native watermarking and ownership verification for 3D Gaussian Splatting assets.
About
I am a Ph.D. student in the College of Computer Science and Technology at
Zhejiang University. My research focuses on 3D vision, 3D Gaussian Splatting,
3D-AIGC, and multi-agent systems, with broader interests in self-supervised
representation learning and embodied visual content creation.
Education
Ph.D. in Artificial Intelligence
2023-09-01
2027-06-30
Zhejiang University
M.S. in Pattern Recognition
2020-09-01
2023-06-30
Shandong University
B.Eng. in Automation
2016-09-01
2020-06-30
Shandong University
Research Interests
3D Vision
3D Gaussian Splatting
3D AI-Generated Content (3D-AIGC)
Multi-agent Systems
Self-supervised Learning
Research
I study 3D vision and generative 3D content creation, especially methods built around 3D Gaussian Splatting. My recent work explores native 3D Gaussian editing, robust 3D asset protection, image-to-3D generation, and agentic systems that can plan, revise, and execute complex editing instructions in 3D scenes.
I also work on self-supervised representation learning for visual and skeleton-based understanding, including contrastive learning, action recognition, gait recognition, and efficient perception models.
I am always interested in collaboration around 3D-AIGC, interactive content editing, and learning systems that connect perception, generation, and controllable creation.
Selected Publications
Variation-aware Flexible 3D Gaussian Editing
Native, flexible 3D Gaussian editing with variation-aware control.
Distilling Multi-view Diffusion Models into 3D Generators
Fast single-image-to-3D Gaussian generation via multi-view diffusion distillation.
Publications
(2026).
NGS-Marker: Robust Native Watermarking for 3D Gaussian Splatting.
ICLR 2026.
(2026).
Placing Any Object at Any 3D Position.
AAAI 2026.
(2026).
Variation-aware Flexible 3D Gaussian Editing.
ICLR 2026.
Projects
Interactive 3D Gaussian Editing
Native and agentic 3D Gaussian editing methods for flexible, multi-turn, and spatially coherent 3D scene manipulation.
•
1 min read
3D Gaussian Generation
Fast and controllable 3D asset generation with 3D Gaussian Splatting, diffusion distillation, and single-image-to-3D pipelines.
•
1 min read
3D-aware Visual Composition
Methods that connect image generation, depth, object identity, and 3D spatial constraints for controllable visual composition.
•
1 min read
Self-supervised Representation Learning
Semantic-aware and progressive self-supervised learning for action recognition, contrastive learning, and visual representation learning.
•
1 min read