Question

9. (10 pts) Consider a convolutional neural network whose inputs are an RGB color image of size 39 x 39 (i.e., the input volume size is 39 x 39 x 3). The network has three convolutional layers and the parameters of each convolutional layer are given in Figure 1. Compute the output volume size of the feature maps at each convolutional layer. Conv1 Conv2 Conv3 Filter size = 3 X 3 # filters = 10 Stride = 1 Padding = 0 Filter size = 5 x 5 # filters = 20 Stride = 2 Padding = 0 Filter size = 5 x 5 # filters = 40 Stride = 2 Padding = 0 Input: 39 x 39 x 3 Figure 1. CNN model.

Answer 1

--> The Output size is calculated using the formula :

      $o = \frac{(i - k + 2*p)}{s} + 1$

     Where,

     o * o is the output size

     i * i is the input size

     k * k is the filter size

     p is padding

     s is stride

--> From the above formula,

-->   Convolution 1 output is :

        o = (39 - 3 + 2*0) / 1 + 1

           = 37

        Size with no of filters 10 is : 37 * 37 * 10

--> Convolution 2 output is :

        o = (37 - 5 + 2*0) / 2 + 1

           = 17

       Size with no of filters 20 is : 17 * 17 * 20

--> Convolution 3 output is :

        o = (17 - 5 + 2*0) / 2 + 1

           = 7

       Size with no of filters 40 is : 7 * 7 * 40