Neural rendering methods represent a crucial and challenging task in computer vision. Recent advancements in 3D Gaussian Splatting have set new benchmarks in rendering quality and speed by combining the strengths of geometric primitives and volumetric representations. In contrast to the point-based approach of 3D Gaussian Splatting, Scaffold-GS obtains anchors to achieve significant memory reduction. In this paper, we use Scaffold-GS as the baseline to achieve high quality rendering with minimal anchors. We propose a depth guide structured 3D Gaussian method for real-time rendering, called DS-GS. Our method employs depth priors to guide two types of anchor growth, ensuring that anchors develop in the correct positions during training. And we implement a differentiable depth rasterizer to enhance the similarity between rendered depth and estimated depth. Additionally, we use multi-scale training to improve rendering quality. We demonstrate the advantages and potential of our approach across a variety of scenarios, effectively reducing anchor redundancy and improving rendering quality. For intance, in the Mip-NeRF360 dataset, we increased the PSNR by 0.6 and reduced the storage size by 30%. Furthermore, our approach shows robustness in experiments with real-world scenes.

Stylized animation is often adored for its innovative and daring visual creativity. Due to the strong visual impact and color contrast inherent in woodcut style design, it has been applied in animation and comics. However, traditional woodcuts, hand-drawn, and previous computer-aided methods have yet to address the issues of dwindling design inspiration, lengthy production times, and complex adjustment procedures. To tackle these challenges, we propose a novel network framework, the Woodcut-style Design Assistant Network (WDANet). Notably, our research is the first to utilize diffusion models to streamline the woodcut-style design process. We curate the Woodcut-62 dataset, which features works from 62 renowned historical artists, to train WDANet in absorbing and learning the aesthetic nuances of woodcut prints, offering users a wealth of design references. Our WDANet integrates text and woodcut-style image features based on a denoising network. WDANet allows users to input or slightly modify a text description to quickly generate accurate, high-quality woodcut-style designs, saving time and offering flexibility. As confirmed by user studies, quantitative and qualitative analyses show that WDANet outperforms the current state-of-the-art in generating woodcut-style images and proves its value as a design aid.

Although virtual sandboxes are an essential component of platforms such as smart cities and smart parks, the efficiency of visual rendering and Web access is frequently hampered by many scenarios and models. This paper combines the polygon reduction algorithm with human visual system characteristics, proposes a real-time adjustment method for model data for virtual scenes, determines the degree of surface reduction based on the distance of the model’s center point in real-time observation, and then adjusts the number of triangular faces of the model in the scene using the collapse principle, to dynamically adjust the data amount of the model. It also uses coroutines to realize the parallel surface reduction of multiple objects and improve the rendering efficiency further. Experiments show that the optimized algorithm does not affect the visual observation effect after reducing the visual distance and storage space by 12%, indicating that it is a highly adaptive virtual scene data optimization method.

The rapid growth of the live streaming industry brought about the VTuber trend, where content creators use character avatars to stream. One of the most accessible way to move a character in real time is by using a vision-based facial motion capture technique. However, previous works still suffer from jittering issues, which hinder the quality of the character’s movements. This work aims to develop a smoothed facial motion capture system that works with a live2D VTuber model. The system combines Mediapipe Face Mesh and OpenCV solutions to capture facial landmarks, which are then used to calculate head pose estimation using the Perspective-n-Point (PnP) function. In addition, the system uses EAR and MAR functions to detect facial features. The motion values obtained from this process are then filtered using a Kalman filter. Finally, the filtered motion data is sent to the Unity engine, which drives the Live2D VTuber model by adjusting the character’s motion parameters. The developed system successfully captures and drives the Live2D VTuber model with smoother motion, overcoming the jitter problem prevalent in previous facial motion capture approaches. The system’s improved motion capture quality makes it a more viable option for a wide range of potential uses.

The work description of virtual reality technology is superior to static reference pictures and videos, which can improve the predictability of decision makers and become a powerful performance tool for urban lighting design. The urban scene is constructed by Unity 3D, and the intelligent lamp pole and Rhododendron landscape lawn lamp are taken as the research objects. On the basis of maximizing the application effect of lamps with the HTC VIVE virtual headset, the city and lamps are digitally mapped 1:1, After applying the model material self-luminous, setting the type of light source and the position of the light source point, setting the transparency mixing, adding glow, and changing the light effect, a highly restored lamp and lamp simulation model are formed. The research results show that this method provides a convenient research method for the three-dimensional reconstruction and visual simulation of urban landscape lamps. It can make decision makers have a clear understanding of the application effect, application quantity and interval of lamps. Immersive observation experience can make the design scheme more honest and reliable. The efficient simulation of lighting lamps will play an important role in the dynamic development of smart cities in the future.