It seems like you used the wrong image for VGG16 and VGG19 architecture. VGG16 and VGG19 have 13 and 16 convolutional layers respectively.