Image Forensic

IML-MPU is a large-scale synthetic forgery dataset generating by Photoshop scripting (PS-Scripting), using a diverse set of samples sourced from VISION, KCMI, and own photographs selected as original images. The dataset comprises three distinct subsets, namely copy-move, splicing, and removal, consisting of 38,000, 43,000, and 32,000 images, respectively. It contains uncompressed TIFF images and JPEG images with different compression factors.

The NANet Dataset is a synthetic dataset specifically designed for splicing forgery detection, using the MSCOCO datasets as its foundational sources. There are 113,964 forgery images in NANet. In this dataset, the positively labeled instances correspond to tampered regions, whereas the negatively labeled instances pertain to non-tampered regions. A pair of positive and negative region belongs to the same category to ensures a reasonable semantic relevance of tampered images within the NANet dataset.

The SPANet Dataset is a synthetic dataset specifically designed for copy-move forgery detection, using the SUN and MSCOCO datasets as its foundational sources. The dataset consists of 550 basic forgery images, including a diverse range of 12 attack types within rotation and scaling variations, and 11 post-processing techniques. Therefore, the number of forgery images in SPANet Dataset is 550 * 12 * 11 = 72600.