Deepfake detection machine learning models
<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-InterpretableRationales for legal investigation purposes</a></p><p dir="ltr">The zip archive contains four trained machine learning models, which are trained to detect deepfake images in their original form, their error level analysis (ELA) processed images, noise analysis (NA) processed images and principal component analysis (PCA) processed images.</p><p dir="ltr">The models were created using Python 3.9 and Tensorflow library version 2.16.2. The architecture of the models are:</p><table><tr><td><p dir="ltr"><b>Layer</b></p></td><td><p dir="ltr"><b>Output Shape</b></p></td></tr><tr><td><p dir="ltr">Input layer</p></td><td><p>180, 180, 3</p></td></tr><tr><td><p dir="ltr">Rescaling</p></td><td><p>180, 180, 3</p></td></tr><tr><td><p dir="ltr">Xception</p></td><td><p>6, 6, 2048</p></td></tr><tr><td><p dir="ltr">Global average pooling 2D</p></td><td><p>2048</p></td></tr><tr><td><p dir="ltr">Dropout</p></td><td><p>2048</p></td></tr><tr><td><p dir="ltr">Dense</p></td><td><p>2</p></td></tr></table><p><br></p>
