Hi guys
for this sliding window explanation I just wanna know why is it the yellow boxes for the first 2 convo layer are of n_W and n_H plus 2 but after max pool just plus 1? is it because this is just for the example or is there something I missed that helps to determine the amt of slide to the convo layer.
Hello @zheng_xiang1
The slide is from C4 W3 Video “Convolutional Implementation of Sliding Windows”.
Yes - 16 x 16 x 3 is for the sake of example. The key is that the example has to be larger than the model’s input (14 x 14) for the purpose of discussing sliding window.
Max-pooling takes the maximum value in every 2x2 slice. The stride for that maxpooling is supposed to be 2, so it takes the max. value in the first 2x2 slice, then skip one slice to the 3rd, and take the max. value in the 3rd slice, and so on and so forth. This process leaves us 6 x 6 max. values.
Raymond
1 Like
