Skip to main content

Table 6 The accuracy of double JPEG compression detection using different training set sizes

From: Double JPEG compression forensics based on a convolutional neural network

QF 2/total number of blocks

33,000

66,000

132,000

264,000

396,000

60

0.500

0.695

0.702

0.728

0.729

80

0.500

0.773

0.780

0.791

0.791