Computer Graphics | Ziqi Lu

WaterLOD

Mon, 09 Jun 2025 00:00:00 +0000

With the progress of computer-aided simulation research, there’s an increasing demand for large-scale 3D fluid simulation in geotechnical and hydraulic fields. These fields require processing large-scale, high-precision fluid simulation data. However, traditional visualization technologies have limitations including low efficiency, poor reality and lack of intuitiveness. At the same time, the promising application prospect of currently popping technologies such as Virtual Reality and Augmented Reality require development of photo-realistic scene fluid visualization.

Therefore, this study concentrates on the photo-realistic scene visualization of large-scale fluid particle data, aiming to rapidly and interactively visualize particle data at around hundred-million level. This study first introduces the continuous level of detail technique into the field of fluid particle data through in-depth observation of input data characteristics to reduce performance waste; At the same time, by monitoring hardware performance, intelligent preloading of future frames of fluid animation can improve the smoothness of video rendering. The core of this study is to develop on the vtkOpenGLFluidMapper class in VTK’s OpenGL extension. By applying the continuous level-of-detail technology and the future frame pre-loading and transmission optimization technique, it establishes an efficient fluid particle data scene visualization method, and implemented a prototype system to validate the feasibility and superiority of the algorithm. The research improves the efficiency of authenticity visualization without significantly compromising visual effects.

FlameGS

Sun, 08 Sep 2024 00:00:00 +0000

FlameGS

July 2024 - Sept. 2024

Research Intern at University of Utah, Advisor:Yang Yin

Developed a comprehensive process pipeline to reconstruct photo-realistic facial meshes, textures, and animations from monocular or multi-camera video sources, enhancing the avatar realism and versatility in various applications
Integrated the FLAME parametric differentiable face model with the Gaussian Splatting method to efffficiently capture detailed features under extreme data distributions
Formulated a transfer algorithm for further simulation on the mixed avatar representation of meshes and gaussians, which are fully controllable and user-friendly in terms of expression, pose, and viewpoint
Achieved comparable image similarity and multi-view consistency with state-of-the-art methods, while also producing more temporally smooth pose and expression animations

Copyright

Header image copyright: by Technical University of Munich

Micro-PT

Tue, 28 Jun 2022 00:00:00 +0000

Features

Distributed Path Tracing / Stochastic Progressive Photon Mapping
OpenMP Multi-threading
Modern Microfacet Material Support
Bounding Volume Hierarchy Acceleration