When I train a neural network in Caffe,I found the blob shape change after the forward propagation

Before I begin the forward propagation,the shape of the blob['Conv_1'] is [1,64,380,576],but after the forward progagation the shape of the blob['Conv_1'] is [1,64,370,576],the parameters of the Conv layer is that{padding = 1,stride = 1,Kersize = 3},I want to know why this condition happens?