caffe 里deconvolution layer （解卷积层)的理解

layer {
name: "deconv5_2"
type: "Deconvolution"
bottom: "conv5_2"
top: "deconv5_2"
param {
lr_mult: 1
decay_mult: 0
}
param {
lr_mult: 2
decay_mult: 0
}
convolution_param {
num_output: 512
pad: 1  #is right?
stride: 1  #is right?
kernel_size: 6  #is right?
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0
}
}
}
output_size = (input_size - 1) * stride + kernel_size - 2 * pad