PERSE
收藏资源简介:
PERSE数据集由首尔国立大学的研究团队创建,旨在生成个性化的3D生成头像。该数据集包含大规模合成的2D视频,每个视频展示了面部表情和视角的一致性变化,并结合了特定面部属性的变化。数据集的创建过程涉及从单张肖像图像生成高质量、逼真的2D视频,并通过3D高斯溅射技术学习连续且解耦的潜在空间。该数据集的应用领域包括虚拟现实和增强现实环境中的个性化头像生成,旨在解决用户希望在虚拟世界中调整面部特征同时保留核心身份的需求。
The PERSE dataset was developed by a research team from Seoul National University, with the objective of generating personalized 3D avatars. This dataset includes large-scale synthesized 2D videos, where each video displays consistent variations in facial expressions and viewpoints, alongside changes in specific facial attributes. The dataset creation process entails generating high-quality, photorealistic 2D videos from a single portrait image, and learning a continuous and disentangled latent space through 3D Gaussian Splatting technology. Its application areas cover personalized avatar generation in virtual reality (VR) and augmented reality (AR) environments, aiming to address the need for users to adjust facial features while retaining their core identity in virtual worlds.




