Fig 3. Operation in a convolutional layer with ReLU and Pooling.

The first column shows a (6 × 6) input example with 3 channels. The columns from second to forth show the interaction of this input with a receptive field (3 × 3, represented in the third column) generating a (6 × 6) output with 1 channel in the forth column. The second column shows a padding with value 1, and the applied stride is also 1. The fifth column represents the output of a ReLU operation. The last column presents the 3 × 3 output for a 2 × 2 window Max Pooling operation (represented in the sixth column).