Fig. 16From: Mobile authentication of copy detection patternsThe second scenario results’ visualization: the 2D distribution of (i) the symbol-wise Hamming distance between the original digital templates \({\textbf {t}}\) and the corresponding estimations \(\hat{{\textbf {t}}}\) obtained via the encoder model trained with respect to the \(\mathcal {D}_{{\textrm{t}} \hat{\textrm{t}}}\) term and (ii) the corresponding responses of the discriminator model trained with respect to the \(\mathcal {D}_{\textrm{t}}\) termBack to article page