Skip to main content

Table 5 Validation accuracy for various validation sets after training on the composite printers data set

From: Printing and scanning investigation for image counter forensics

 

Bayar2016

Xception

Proposed model

Dell (4c)

0.6506

0.649

0.713

Xerox1 (4c)

0.7001

0.626

0.696

Xeros2 (4c)

0.5381

0.623

0.663

JPEG (4c)

0.2643

0.2902

0.2601

Original (4c)

0.2617

0.2847

0.2449

  1. Bold values refer to models that perform better than the rest - to highlight the model performance - its a common practice to do this and usually helps improve readability
  2. One possible explanation for the poor validation accuracy on a single printer could be the small size of the data set. To investigate this, we combine the images from all three printers for training, but note that performance on a single printer does not improve. Here 4c indicates that we used the restricted set of manipulations (AWGN, GB, MF, and PR) (see Sect. 4)