On July 31, the CCF-Tencent Rhino-Bird Fund announced the list of scholars selected for its 2026 program, Assistant Professor Dr. Fangneng Zhan of the Division of Arts and Machine Creativity (AMC), was selected as a recipient of the 2026 CCF-Tencent Rhino-Bird Fund, joining 370 outstanding scholars from leading universities and research institutions.
Project details
Dr. Zhan Fangneng’s project, “3D Geometry-Prior-Driven Spatiotemporally Consistent Video Generation for Free-Viewpoint Complex Scenes,” seeks to address key challenges in generative video technology by enabling the creation of realistic free-viewpoint videos from a single image or a small number of reference images, while maintaining stable 3D structure and long-term temporal consistency across large viewpoint changes.
The project will leverage a frozen 4D geometry foundation model as a teacher during training, distilling rich geometric priors, including depth, point clouds, camera motion, scene flow, visibility, and point trajectories, into a video Diffusion Transformer. By incorporating these geometric cues, the model is expected to achieve greater spatial coherence and more accurate scene dynamics across diverse viewing angles.
At inference time, only a streamlined single-stream video generator will be retained, allowing the system to enhance geometric fidelity and cross-view consistency without significantly increasing computational cost. This design is intended to balance generation quality with practical deployment efficiency.
Beyond algorithm development, the research team will establish a benchmark for complex-scene video generation and develop evaluation metrics covering depth consistency, point-tracking accuracy, reprojection error, geometric drift, and hallucinations in previously unseen regions. The benchmark is expected to provide a standardized framework for assessing geometric and temporal quality in next-generation video generation systems.
The project’s outcomes could support a wide range of applications, including XR content creation, 3D and 4D product visualization, film and game pre-visualization, and embodied AI world simulation, contributing to the advancement of immersive digital content production and intelligent virtual environment generation.
About the CCF-Tencent Rhino-Bird Fund
The Rhino-Bird Fund is dedicated to promoting deep industry-academia collaboration by connecting pioneering academic research with real-world industrial needs. As AI technologies continue to evolve rapidly, the fund brings together outstanding scholars and industry experts to explore new frontiers, advance technological innovation, and create sustainable social value through responsible AI development.
This year, the fund introduced 43 research topics across seven strategic areas, including foundation models, autonomous and multi-agent systems, and trusted data construction. Through the Rhino-Bird Scholar Program, the fund supports the development and deployment of frontier AI research in practical applications, helping bridge the gap between academic discovery and industry innovation.
