PetroGAN

PetroGAN是一个基于生成对抗网络（GAN）的合成岩石化石数据集，由国立哥伦比亚大学地质科学系开发。该数据集包含10070张岩石薄片图像，用于训练GAN模型以生成高质量的合成岩石化石图像。数据集的创建过程涉及使用StyleGAN2架构进行图像的统计和美学特征的复制，并通过专家评审确保生成图像的真实性。PetroGAN的应用领域主要集中在地质科学中，旨在解决传统地质数据集获取成本高、时间消耗大的问题，同时支持深度学习算法在图像分析中的应用。

PetroGAN is a synthetic rock fossil dataset developed by the Department of Geological Sciences of the National University of Colombia, based on generative adversarial networks (GANs). This dataset includes 10,070 thin-section rock images, and is designed to train GAN models for generating high-quality synthetic rock fossil images. The creation of this dataset involves adopting the StyleGAN2 architecture to replicate the statistical and aesthetic features of the images, with expert reviews carried out to verify the authenticity of the generated images. PetroGAN is mainly applied in geological sciences, aiming to solve the problems of high acquisition costs and time-intensive nature of traditional geological datasets, while supporting the application of deep learning algorithms in image analysis.