ibm-research/AIX360-CEM-MAF-data
收藏资源简介:
该数据集用于AIX360库中的CEM-MAF(对比解释方法-多属性翻转)示例,与Jupyter Notebook CEM-MAF example 对应。数据包括每个图像的三类文件:.png图像文件、用于查看图像的.npy文件,以及包含潜在特征的.npy文件,这些特征可通过适当的生成对抗网络(GAN)生成图像。图像基于在CelebA名人脸数据集上训练的GAN生成,CelebA数据集来自Liu等人2015年ICCV论文中的研究。数据集主要用于机器学习和解释性AI应用,如人脸属性分析和模型解释。
This dataset is designed for the CEM-MAF (Contrastive Explanation Method - Multi-Attribute Flip) example in the AIX360 library, corresponding to the Jupyter Notebook titled "CEM-MAF example". It contains three types of files for each image: .png image files, .npy files for image visualization, and .npy files containing latent features that can generate images via appropriate Generative Adversarial Networks (GANs). The images are produced by a GAN trained on the CelebA celebrity face dataset, which originates from the research published in the 2015 ICCV paper by Liu et al. This dataset is primarily applied to machine learning and interpretable AI scenarios, including facial attribute analysis and model explainability.




