2-D convolution. Given the input and two kernels as shown in Fig. 10.43,

a. compute the feature maps by applying kernel $K_1,K_2$ (stride is defined as 1 , and zeros are padded around $I$ ), respectively;

b. re-compute the feature maps if the stride is 2 ;

c. compute the new feature maps after we apply a downsampling process, max pooling, and average pooling (filter size is $2\times 2$ ), respectively, to the obtained feature maps in (a).

Transcribed Image Text:

FIGURE 10.43 0 0 1 1 101010 1101001 0010010 1 0 1 0 1 1 Input / 1 001 100 10 11 0 1 1 0 0 1 0 1 0 1 0 1 0 1 0 0 1 1 0 0 0 1 0 Kernel 0 1 0 K 0 0 1 1 11 1 K 0 1 10 1 1 01