The PS-Battles dataset is collected from a large-scale image processing enthusiast community and provides a basis for media derivation and manipulation detection in the field of vision. The dataset consists of 102,028 images and is divided into 11'142 subsets, each of which contains original images and a different number of processed images.