Figure 5
The diagram shows a left-to-right layered architecture beginning with an “Input Layer” on the far left containing a vertical stack of rectangular cells labeled X 1, X 2, ellipsis, and X 39. A rightward arrow from this pointing to two sequential “Conv-1 D” blocks, each labeled “32 filters, Kernel equals 3”, shown as stacked rectangular filter icons with numbers 1 through 32 and small dot markers. A rightward arrow from this leads to a “Max Pooling” layer represented by overlapping rectangular blocks, then a rightward arrow to a “Dropout” layer labeled “50 percent”, depicted as semi-transparent stacked rectangles. Then a rightward arrow enters an inception block enclosed by a bracket labeled “Inception Block”, where three parallel “Conv-1 D” paths branch vertically with arrows: the top path labeled “32 filters, Kernel equals 1”, the middle path labeled “32 filters, Kernel equals 3”, and the bottom path labeled “32 filters, Kernel equals 5”, each shown with stacked filter icons numbered 1 through 32. All three paths direct arrows to a circular merge symbol with a cross labeled “Concatenate”, followed by a rightward arrow to a tall vertical rectangular block labeled “Inception Output”, then a rightward arrow to another “Max Pooling” layer shown as stacked rectangles, followed by a rightward arrow to a “Dropout” layer labeled “30 percent”. Then a rightward arrow to a dense section labeled “2 Fully Connected Layers” represented by a vertical rectangular block and a network of circular nodes connected by lines, followed by a rightward arrow to an “Output Layer” labeled “Global Average Pooling”, depicted as a vertical rectangle with two filled circular nodes. A final rightward arrow pointing to text reading “D D O S or Benign”?, with bottom brackets labeling sections as “Input Layer”, “Initial Convolutional Layers”, “Inception Block”, “Dense Layers”, and “Output Layer”.The architecture of the proposed CNN model. Source: Created by the authors