Fig. 3From: Cancelable templates for secure face verification based on deep learning and random projectionsExamples of pre-images generated with a NbNet model (trained on \(r=128\) reconstructed embeddings) from reconstructed embeddings with different projection sizes (third to fifth column) and from clean embeddings taken at the output of FeatureNet+ (second column), compared against the original image in the pixel domain (first column)Back to article page