<table><tr><td><p dir="ltr">This dataset is related to the following project:</p><p dir="ltr"><a href="https://doi.org/10.21954/ou.rd.c.8146037" rel="noreferrer" target="_blank">Explainable Deepfake Detection: A Multi-Model Framework with Human-Interpretable Rationales for legal investigation purposes</a></p><p dir="ltr">This dataset contains the test images used to evaluate our deepfake detection framework. It originally contained 20,000 real and deepfake images, but as some 2600 files are protected by the UK Crown and we do not have a permission to reproduced them, so these files were removed.</p><p dir="ltr">Our framework contains 4 machine learning models, which feed in the original images, error-level analysis (ELA) images, noise analysis (NA) images and Principal Component Analysis (PCA) images. The models were created using Tensorflow version 2.26.2.</p><p dir="ltr">The zip archive contains the original, ELA, NA and PCA images in separate folders.</p></td></tr></table><p><br></p>