The image sources of FakeClue include the FF++ and GenImage datasets. I would like to confirm a detail regarding the data sampling scope of these two datasets: were the images selected from their commonly used training subsets, namely the TrainSet of FF++ and the SD1.4 subset of GenImage, or were they sampled from the entirety of the two datasets respectively?
作者您好，感谢开源数据集。我看到FakeClue中的图像来源包含了FF++和GenImage。我想问问关于这两个数据集，是从中常用的训练子集中选取了图片（FF++的TrainSet和GenImage的SD1.4），还是从整个数据集中选取了图片？
The image sources of FakeClue include the FF++ and GenImage datasets. I would like to confirm a detail regarding the data sampling scope of these two datasets: were the images selected from their commonly used training subsets, namely the TrainSet of FF++ and the SD1.4 subset of GenImage, or were they sampled from the entirety of the two datasets respectively?
作者您好，感谢开源数据集。我看到FakeClue中的图像来源包含了FF++和GenImage。我想问问关于这两个数据集，是从中常用的训练子集中选取了图片（FF++的TrainSet和GenImage的SD1.4），还是从整个数据集中选取了图片？