Notes—Convolutional layer

(this img is made by myself)

note:
Convolutional layer:
convolve the filter with image,slide over the image spatially,computing dot products.

stride:step

zerp-padding:
(1)reserve feature map’s resolution
(2)better utilize border information,which improves performance

(this img is downloaded from internet)