Synthetic Dataset from GANs
收藏arXiv2025-09-30 收录
下载链接：
https://haoran-wei.github.io/Portrait-Diffusion
下载链接
链接失效反馈官方服务：
资源简介：
该数据集是为了训练混合先验扩散模型（HPDM）以生成3D人像而设计的合成数据集。它采用了生成对抗网络（GAN）方法进行生成，并包含了从360°的水平视角以及从-30°到30°的俯仰角的不同视角的训练数据。该数据集的任务是用于3D人像的生成。
This is a synthetic dataset developed for training Hybrid Prior Diffusion Models (HPDM) to generate 3D human portraits. It leverages Generative Adversarial Network (GAN) techniques for data generation, and encompasses training data with diverse viewing angles, including a 360° horizontal field of view and a pitch angle range spanning from -30° to 30°. The core task of this dataset is 3D human portrait generation.
