Figure 5. Overview of the proposed

Figure 5. Overview of the proposed Fully Convolutional Neural Network architecture. The neural network consist of three convolutional layers. First two convolutional layers contain 32 filters each with size

3×3

3×3

each. The last layer replaces the fully connected layer with a convolutional layer containing three filters with size

1×1

1×1

and a softmax regression function to obtain class label probability for each pixel.