一个预测层的网络结构如下所示：
可以看到，是由三个分支组成的，分别是"PriorBox"层，以及conf、loc的预测层，其中，conf与loc的预测层的参数是由PriorBox的参数计算得到的，具体计算公式如下：
min_size与max_size分别对应一个尺度的预测框（有几个就对应几个预测框），in_size只管自己的预测，而max_size是与aspect_ratio联系在一起的；
filp参数是对应aspect_ratio的预测框*2，以几个max_size，再乘以几；最终得到结果为A
conf、loc的参数是在A的基础上再乘以类别数（加背景），以及4
如下，是需要预测两类的其中一个尺度的网络参数；
如上算出的是，每个格子需要预测的conf以及loc的个数；
每个预测层有H*W个格子，因此，总共预测的loc以及conf的个数是需要乘以H*W的；
如下是某一个层的例子（转自：http://www.360doc.com/content/17/1013/16/42392246_694639090.shtml）
注意最后这里的num_priorbox的值与前面的并不一样，这里是每个预测层所有的输出框的个数：
layer {
name: "combined_2_EltwisePROD_relu"
type: "ReLU"
bottom: "combined_2_EltwisePROD"
top: "combined_2_EltwisePROD_relu"
}
###########################################
###################################################################
layer {
name: "rescombined_2_EltwisePROD_relu_inter256_mbox_locnew_inter"
type: "Convolution"
bottom: "combined_2_EltwisePROD_relu"
top: "rescombined_2_EltwisePROD_relu_inter256_mbox_locnew_inter"
param {
lr_mult:
decay_mult:
}
convolution_param {
num_output:
bias_term: false
pad:
kernel_size:
stride:
weight_filler {
type: "gaussian"
std: 0.01
}
}
}
layer {
name: "rescombined_2_EltwisePROD_relu_inter256_mbox_locnew_inter_bn"
type: "BatchNorm"
bottom: "rescombined_2_EltwisePROD_relu_inter256_mbox_locnew_inter"
top: "rescombined_2_EltwisePROD_relu_inter256_mbox_locnew_inter"
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
batch_norm_param {
moving_average_fraction: 0.999
eps: 0.001
}
}
layer {
name: "rescombined_2_EltwisePROD_relu_inter256_mbox_locnew_inter_scale"
type: "Scale"
bottom: "rescombined_2_EltwisePROD_relu_inter256_mbox_locnew_inter"
top: "rescombined_2_EltwisePROD_relu_inter256_mbox_locnew_inter"
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
scale_param {
filler {
type: "constant"
value: 1.0
}
bias_term: true
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "rescombined_2i_EltwisePROD_relu_inter256_mbox_locnew_inter"
type: "Convolution"
bottom: "combined_2_EltwisePROD_relu"
top: "rescombined_2i_EltwisePROD_relu_inter256_mbox_locnew_inter"
param {
lr_mult:
decay_mult:
}
convolution_param {
num_output:
bias_term: false
pad:
kernel_size:
stride:
weight_filler {
type: "gaussian"
std: 0.01
}
}
}
layer {
name: "rescombined_2i_EltwisePROD_relu_inter256_mbox_locnew_inter_bn"
type: "BatchNorm"
bottom: "rescombined_2i_EltwisePROD_relu_inter256_mbox_locnew_inter"
top: "rescombined_2i_EltwisePROD_relu_inter256_mbox_locnew_inter"
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
batch_norm_param {
moving_average_fraction: 0.999
eps: 0.001
}
}
layer {
name: "rescombined_2i_EltwisePROD_relu_inter256_mbox_locnew_inter_scale"
type: "Scale"
bottom: "rescombined_2i_EltwisePROD_relu_inter256_mbox_locnew_inter"
top: "rescombined_2i_EltwisePROD_relu_inter256_mbox_locnew_inter"
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
scale_param {
filler {
type: "constant"
value: 1.0
}
bias_term: true
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "combined_2_EltwisePROD_relu_mbox_loc"
type: "Convolution"
bottom: "rescombined_2_EltwisePROD_relu_inter256_mbox_locnew_inter"
top: "combined_2_EltwisePROD_relu_mbox_loc"
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
convolution_param {
engine: CAFFE
num_output:
pad:
kernel_size:
stride:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value:
}
}
}
layer {
name: "combined_2_EltwisePROD_relu_mbox_loc_perm"
type: "Permute"
bottom: "combined_2_EltwisePROD_relu_mbox_loc"
top: "combined_2_EltwisePROD_relu_mbox_loc_perm"
permute_param {
order:
order:
order:
order:
}
}
layer {
name: "combined_2_EltwisePROD_relu_mbox_loc_flat"
type: "Flatten"
bottom: "combined_2_EltwisePROD_relu_mbox_loc_perm"
top: "combined_2_EltwisePROD_relu_mbox_loc_flat"
flatten_param {
axis:
}
}
layer {
name: "combined_2_EltwisePROD_relu_mbox_conf_new"
type: "Convolution"
bottom: "rescombined_2i_EltwisePROD_relu_inter256_mbox_locnew_inter"
top: "combined_2_EltwisePROD_relu_mbox_conf_new"
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
convolution_param {
engine: CAFFE
num_output:
pad:
kernel_size:
stride:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value:
}
}
}
layer {
name: "combined_2_EltwisePROD_relu_mbox_conf_new_perm"
type: "Permute"
bottom: "combined_2_EltwisePROD_relu_mbox_conf_new"
top: "combined_2_EltwisePROD_relu_mbox_conf_new_perm"
permute_param {
order:
order:
order:
order:
}
}
layer {
name: "combined_2_EltwisePROD_relu_mbox_conf_new_flat"
type: "Flatten"
bottom: "combined_2_EltwisePROD_relu_mbox_conf_new_perm"
top: "combined_2_EltwisePROD_relu_mbox_conf_new_flat"
flatten_param {
axis:
}
}
layer {
name: "combined_2_EltwisePROD_relu_mbox_priorbox"
type: "PriorBox"
bottom: "combined_2_EltwisePROD_relu"
bottom: "data"
top: "combined_2_EltwisePROD_relu_mbox_priorbox"
prior_box_param {
min_size: 12.0
min_size: 6.0
max_size: 30.0
max_size: 20.0
aspect_ratio:
aspect_ratio: 2.5
aspect_ratio:
flip: true
clip: false
variance: 0.1
variance: 0.1
variance: 0.2
variance: 0.2
step:
offset: 0.5
}
}