A5082525759- Computer Graphics and Visualization Techniques
- Advanced Vision and Imaging
- Video Coding and Compression Technologies
- 3D Shape Modeling and Analysis
- Robotics and Sensor-Based Localization
- Generative Adversarial Networks and Image Synthesis
- Human Motion and Animation
- Advanced Image and Video Retrieval Techniques
Shanghai Jiao Tong University
2022-2024
Depth image-based rendering (DIBR) view synthesis is the most widely employed method in real-time FVV research. Despite recent progress, DIBR-based approaches are not sufficiently simple and effective filling holes artifacts. Additionally, they use RGB-D cameras, which difficult to adopt or take considerable time estimate high-quality depth images. This paper introduces a system based on DIBR estimation network. includes 12-view synchronous camera system, new multistage network,...
360° panoramas are extensively utilized as environmental light sources in computer graphics. However, capturing a × 180° panorama poses challenges due to the necessity of specialized and costly equipment, additional human resources. Prior studies develop various learning-based generative methods synthesize from single Narrow Field-of-View (NFoV) image, but they limited alterable input patterns, generation quality, controllability. To address these issues, we propose novel pipeline called...
Recently, human rendering has attracted many attention thanks to its vast applications. With new advances in neural and radiance field, synthesizing realistic novel view images from multi-view camera can be achived with less manual labour. However, due the data-driven nature of such algorithms, efficiency for both time computation unsatisfying. Hence, we propose an efficient pipeline, generating geometric semantic guidances as priors further enhance quality. Specifically, a part parsing...
Free viewpoint video (FVV) has drawn great attention in recent years, which provides viewers with strong interactive and immersive experience. Despite the developments made, further progress of FVV research is limited by existing datasets that mostly have too few number camera views, or static scenes. To overcome limitations, this paper, we present a new dynamic RGB-D dataset up to 12 views. Our consists 13 groups sequences are taken at same scene, group empty scene. Each HD synchronized...
With the increasing popularity of commercial depth cameras, 3D reconstruction dynamic scenes has aroused widespread interest. Although many novel applications have been unlocked, real-time performance is still a big problem. In this paper, low-cost, system: LiveRecon3D, presented, with multiple RGB-D cameras connected to one single computer. The goal system provide an interactive frame rate for content capture and rendering at reduced cost. proposed system, we adopt scalable volume structure...