refinedet网络结构
refinedet只预测4个层,并且只有conv6_1、conv6_2,没有ssd中的conv7、8、9
refinedet的4个层都只有1个aspect ratio和1个min_size,所以每层每个点只有3个anchor,arm中做location的conv4_3_norm_mbox_loc等层都是3*4个channel,做confidence的conv4_3_norm_mbox_conf都是6个channel,因为这里变成了2分类,每个anchor必须要有negative和positive的概率
refinedet是两步都要回归bounding box的框,refinedet中的odm_loss就相当于ssd中的mbox_loss,mbox_loss获得了anchor的坐标后会加上回归再进行训练,odm_loss获得anchor的坐标后先要加上arm_loc的回归,再加odm_loc的回归,这样再去进行loss计算.
name: "vgg_1/8"
layer {
name: "data"
type: "AnnotatedData"
top: "data"
top: "label"
include {
phase: TRAIN
}
transform_param {
mirror: true
mean_value: 104.0
mean_value: 117.0
mean_value: 123.0
resize_param {
prob: 1.0
resize_mode: WARP
height:
width:
interp_mode: LINEAR
interp_mode: AREA
interp_mode: NEAREST
interp_mode: CUBIC
interp_mode: LANCZOS4
}
emit_constraint {
emit_type: CENTER
}
distort_param {
brightness_prob: 0.5
brightness_delta: 32.0
contrast_prob: 0.5
contrast_lower: 0.5
contrast_upper: 1.5
hue_prob: 0.5
hue_delta: 18.0
saturation_prob: 0.5
saturation_lower: 0.5
saturation_upper: 1.5
random_order_prob: 0.0
}
expand_param {
prob: 0.5
max_expand_ratio: 4.0
}
}
data_param {
source:"examples/cityscapes/cityscapes_train_lmdb"
batch_size:
backend: LMDB
}
annotated_data_param {
batch_sampler {
max_sample:
max_trials:
}
batch_sampler {
sampler {
min_scale: 0.300000011921
max_scale: 1.0
min_aspect_ratio: 0.5
max_aspect_ratio: 2.0
}
sample_constraint {
min_jaccard_overlap: 0.10000000149
}
max_sample:
max_trials:
}
batch_sampler {
sampler {
min_scale: 0.300000011921
max_scale: 1.0
min_aspect_ratio: 0.5
max_aspect_ratio: 2.0
}
sample_constraint {
min_jaccard_overlap: 0.300000011921
}
max_sample:
max_trials:
}
batch_sampler {
sampler {
min_scale: 0.300000011921
max_scale: 1.0
min_aspect_ratio: 0.5
max_aspect_ratio: 2.0
}
sample_constraint {
min_jaccard_overlap: 0.5
}
max_sample:
max_trials:
}
batch_sampler {
sampler {
min_scale: 0.300000011921
max_scale: 1.0
min_aspect_ratio: 0.5
max_aspect_ratio: 2.0
}
sample_constraint {
min_jaccard_overlap: 0.699999988079
}
max_sample:
max_trials:
}
batch_sampler {
sampler {
min_scale: 0.300000011921
max_scale: 1.0
min_aspect_ratio: 0.5
max_aspect_ratio: 2.0
}
sample_constraint {
min_jaccard_overlap: 0.899999976158
}
max_sample:
max_trials:
}
batch_sampler {
sampler {
min_scale: 0.300000011921
max_scale: 1.0
min_aspect_ratio: 0.5
max_aspect_ratio: 2.0
}
sample_constraint {
max_jaccard_overlap: 1.0
}
max_sample:
max_trials:
}
label_map_file: "data/cityscapes/labelmap_cityscapes.prototxt"
}
}
layer {
name: "conv1_1"
type: "Convolution"
bottom: "data"
top: "conv1_1"
param {
lr_mult: 1.0
decay_mult: 1.0
}
param {
lr_mult: 2.0
decay_mult: 0.0
}
convolution_param {
num_output:
pad:
kernel_size:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "relu1_1"
type: "ReLU"
bottom: "conv1_1"
top: "conv1_1"
}
layer {
name: "conv1_2"
type: "Convolution"
bottom: "conv1_1"
top: "conv1_2"
param {
lr_mult: 1.0
decay_mult: 1.0
}
param {
lr_mult: 2.0
decay_mult: 0.0
}
convolution_param {
num_output:
pad:
kernel_size:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "relu1_2"
type: "ReLU"
bottom: "conv1_2"
top: "conv1_2"
}
layer {
name: "pool1"
type: "Pooling"
bottom: "conv1_2"
top: "pool1"
pooling_param {
pool: MAX
kernel_size:
stride:
}
}
layer {
name: "conv2_1"
type: "Convolution"
bottom: "pool1"
top: "conv2_1"
param {
lr_mult: 1.0
decay_mult: 1.0
}
param {
lr_mult: 2.0
decay_mult: 0.0
}
convolution_param {
num_output:
pad:
kernel_size:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "relu2_1"
type: "ReLU"
bottom: "conv2_1"
top: "conv2_1"
}
layer {
name: "conv2_2"
type: "Convolution"
bottom: "conv2_1"
top: "conv2_2"
param {
lr_mult: 1.0
decay_mult: 1.0
}
param {
lr_mult: 2.0
decay_mult: 0.0
}
convolution_param {
num_output:
pad:
kernel_size:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "relu2_2"
type: "ReLU"
bottom: "conv2_2"
top: "conv2_2"
}
layer {
name: "pool2"
type: "Pooling"
bottom: "conv2_2"
top: "pool2"
pooling_param {
pool: MAX
kernel_size:
stride:
}
}
layer {
name: "conv3_1"
type: "Convolution"
bottom: "pool2"
top: "conv3_1"
param {
lr_mult: 1.0
decay_mult: 1.0
}
param {
lr_mult: 2.0
decay_mult: 0.0
}
convolution_param {
num_output:
pad:
kernel_size:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "relu3_1"
type: "ReLU"
bottom: "conv3_1"
top: "conv3_1"
}
layer {
name: "conv3_2"
type: "Convolution"
bottom: "conv3_1"
top: "conv3_2"
param {
lr_mult: 1.0
decay_mult: 1.0
}
param {
lr_mult: 2.0
decay_mult: 0.0
}
convolution_param {
num_output:
pad:
kernel_size:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "relu3_2"
type: "ReLU"
bottom: "conv3_2"
top: "conv3_2"
}
layer {
name: "conv3_3"
type: "Convolution"
bottom: "conv3_2"
top: "conv3_3"
param {
lr_mult: 1.0
decay_mult: 1.0
}
param {
lr_mult: 2.0
decay_mult: 0.0
}
convolution_param {
num_output:
pad:
kernel_size:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "relu3_3"
type: "ReLU"
bottom: "conv3_3"
top: "conv3_3"
}
layer {
name: "pool3"
type: "Pooling"
bottom: "conv3_3"
top: "pool3"
pooling_param {
pool: MAX
kernel_size:
stride:
}
}
layer {
name: "conv4_1"
type: "Convolution"
bottom: "pool3"
top: "conv4_1"
param {
lr_mult: 1.0
decay_mult: 1.0
}
param {
lr_mult: 2.0
decay_mult: 0.0
}
convolution_param {
num_output:
pad:
kernel_size:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "relu4_1"
type: "ReLU"
bottom: "conv4_1"
top: "conv4_1"
}
layer {
name: "conv4_2"
type: "Convolution"
bottom: "conv4_1"
top: "conv4_2"
param {
lr_mult: 1.0
decay_mult: 1.0
}
param {
lr_mult: 2.0
decay_mult: 0.0
}
convolution_param {
num_output:
pad:
kernel_size:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "relu4_2"
type: "ReLU"
bottom: "conv4_2"
top: "conv4_2"
}
layer {
name: "conv4_3"
type: "Convolution"
bottom: "conv4_2"
top: "conv4_3"
param {
lr_mult: 1.0
decay_mult: 1.0
}
param {
lr_mult: 2.0
decay_mult: 0.0
}
convolution_param {
num_output:
pad:
kernel_size:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "relu4_3"
type: "ReLU"
bottom: "conv4_3"
top: "conv4_3"
}
layer {
name: "pool4"
type: "Pooling"
bottom: "conv4_3"
top: "pool4"
pooling_param {
pool: MAX
kernel_size:
stride:
}
}
layer {
name: "conv5_1"
type: "Convolution"
bottom: "pool4"
top: "conv5_1"
param {
lr_mult: 1.0
decay_mult: 1.0
}
param {
lr_mult: 2.0
decay_mult: 0.0
}
convolution_param {
num_output:
pad:
kernel_size:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "relu5_1"
type: "ReLU"
bottom: "conv5_1"
top: "conv5_1"
}
layer {
name: "conv5_2"
type: "Convolution"
bottom: "conv5_1"
top: "conv5_2"
param {
lr_mult: 1.0
decay_mult: 1.0
}
param {
lr_mult: 2.0
decay_mult: 0.0
}
convolution_param {
num_output:
pad:
kernel_size:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "relu5_2"
type: "ReLU"
bottom: "conv5_2"
top: "conv5_2"
}
layer {
name: "conv5_3"
type: "Convolution"
bottom: "conv5_2"
top: "conv5_3"
param {
lr_mult: 1.0
decay_mult: 1.0
}
param {
lr_mult: 2.0
decay_mult: 0.0
}
convolution_param {
num_output:
pad:
kernel_size:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "relu5_3"
type: "ReLU"
bottom: "conv5_3"
top: "conv5_3"
}
layer {
name: "pool5"
type: "Pooling"
bottom: "conv5_3"
top: "pool5"
pooling_param {
pool: MAX
kernel_size:
stride:
}
}
layer {
name: "fc6"
type: "Convolution"
bottom: "pool5"
top: "fc6"
param {
lr_mult: 1.0
decay_mult: 1.0
}
param {
lr_mult: 2.0
decay_mult: 0.0
}
convolution_param {
num_output:
pad:
kernel_size:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "relu6"
type: "ReLU"
bottom: "fc6"
top: "fc6"
}
layer {
name: "fc7"
type: "Convolution"
bottom: "fc6"
top: "fc7"
param {
lr_mult: 1.0
decay_mult: 1.0
}
param {
lr_mult: 2.0
decay_mult: 0.0
}
convolution_param {
num_output:
kernel_size:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "relu7"
type: "ReLU"
bottom: "fc7"
top: "fc7"
}
layer {
name: "conv6_1"
type: "Convolution"
bottom: "fc7"
top: "conv6_1"
param {
lr_mult: 1.0
decay_mult: 1.0
}
param {
lr_mult: 2.0
decay_mult: 0.0
}
convolution_param {
num_output:
pad:
kernel_size:
stride:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "conv6_1_relu"
type: "ReLU"
bottom: "conv6_1"
top: "conv6_1"
}
layer {
name: "conv6_2"
type: "Convolution"
bottom: "conv6_1"
top: "conv6_2"
param {
lr_mult: 1.0
decay_mult: 1.0
}
param {
lr_mult: 2.0
decay_mult: 0.0
}
convolution_param {
num_output:
pad:
kernel_size:
stride:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "conv6_2_relu"
type: "ReLU"
bottom: "conv6_2"
top: "conv6_2"
}
layer {
name: "conv4_3_norm_mbox_loc"
type: "Convolution"
bottom: "conv4_3"
top: "conv4_3_norm_mbox_loc"
param {
lr_mult: 1.0
decay_mult: 1.0
}
param {
lr_mult: 2.0
decay_mult: 0.0
}
convolution_param {
num_output:
pad:
kernel_size:
stride:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "conv4_3_norm_mbox_loc_perm"
type: "Permute"
bottom: "conv4_3_norm_mbox_loc"
top: "conv4_3_norm_mbox_loc_perm"
permute_param {
order:
order:
order:
order:
}
}
layer {
name: "conv4_3_norm_mbox_loc_flat"
type: "Flatten"
bottom: "conv4_3_norm_mbox_loc_perm"
top: "conv4_3_norm_mbox_loc_flat"
flatten_param {
axis:
}
}
layer {
name: "conv4_3_norm_mbox_conf"
type: "Convolution"
bottom: "conv4_3"
top: "conv4_3_norm_mbox_conf"
param {
lr_mult: 1.0
decay_mult: 1.0
}
param {
lr_mult: 2.0
decay_mult: 0.0
}
convolution_param {
num_output:
pad:
kernel_size:
stride:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "conv4_3_norm_mbox_conf_perm"
type: "Permute"
bottom: "conv4_3_norm_mbox_conf"
top: "conv4_3_norm_mbox_conf_perm"
permute_param {
order:
order:
order:
order:
}
}
layer {
name: "conv4_3_norm_mbox_conf_flat"
type: "Flatten"
bottom: "conv4_3_norm_mbox_conf_perm"
top: "conv4_3_norm_mbox_conf_flat"
flatten_param {
axis:
}
}
layer {
name: "conv4_3_norm_mbox_priorbox"
type: "PriorBox"
bottom: "conv4_3"
bottom: "data"
top: "conv4_3_norm_mbox_priorbox"
prior_box_param {
min_size: 16.0
aspect_ratio: 2.0
flip: true
clip: false
variance: 0.10000000149
variance: 0.10000000149
variance: 0.20000000298
variance: 0.20000000298
step: 8.0
offset: 0.5
}
}
layer {
name: "conv5_3_norm_mbox_loc"
type: "Convolution"
bottom: "conv5_3"
top: "conv5_3_norm_mbox_loc"
param {
lr_mult: 1.0
decay_mult: 1.0
}
param {
lr_mult: 2.0
decay_mult: 0.0
}
convolution_param {
num_output:
pad:
kernel_size:
stride:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "conv5_3_norm_mbox_loc_perm"
type: "Permute"
bottom: "conv5_3_norm_mbox_loc"
top: "conv5_3_norm_mbox_loc_perm"
permute_param {
order:
order:
order:
order:
}
}
layer {
name: "conv5_3_norm_mbox_loc_flat"
type: "Flatten"
bottom: "conv5_3_norm_mbox_loc_perm"
top: "conv5_3_norm_mbox_loc_flat"
flatten_param {
axis:
}
}
layer {
name: "conv5_3_norm_mbox_conf"
type: "Convolution"
bottom: "conv5_3"
top: "conv5_3_norm_mbox_conf"
param {
lr_mult: 1.0
decay_mult: 1.0
}
param {
lr_mult: 2.0
decay_mult: 0.0
}
convolution_param {
num_output:
pad:
kernel_size:
stride:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "conv5_3_norm_mbox_conf_perm"
type: "Permute"
bottom: "conv5_3_norm_mbox_conf"
top: "conv5_3_norm_mbox_conf_perm"
permute_param {
order:
order:
order:
order:
}
}
layer {
name: "conv5_3_norm_mbox_conf_flat"
type: "Flatten"
bottom: "conv5_3_norm_mbox_conf_perm"
top: "conv5_3_norm_mbox_conf_flat"
flatten_param {
axis:
}
}
layer {
name: "conv5_3_norm_mbox_priorbox"
type: "PriorBox"
bottom: "conv5_3"
bottom: "data"
top: "conv5_3_norm_mbox_priorbox"
prior_box_param {
min_size: 32.0
aspect_ratio: 2.0
flip: true
clip: false
variance: 0.10000000149
variance: 0.10000000149
variance: 0.20000000298
variance: 0.20000000298
step: 16.0
offset: 0.5
}
}
layer {
name: "fc7_mbox_loc"
type: "Convolution"
bottom: "fc7"
top: "fc7_mbox_loc"
param {
lr_mult: 1.0
decay_mult: 1.0
}
param {
lr_mult: 2.0
decay_mult: 0.0
}
convolution_param {
num_output:
pad:
kernel_size:
stride:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "fc7_mbox_loc_perm"
type: "Permute"
bottom: "fc7_mbox_loc"
top: "fc7_mbox_loc_perm"
permute_param {
order:
order:
order:
order:
}
}
layer {
name: "fc7_mbox_loc_flat"
type: "Flatten"
bottom: "fc7_mbox_loc_perm"
top: "fc7_mbox_loc_flat"
flatten_param {
axis:
}
}
layer {
name: "fc7_mbox_conf"
type: "Convolution"
bottom: "fc7"
top: "fc7_mbox_conf"
param {
lr_mult: 1.0
decay_mult: 1.0
}
param {
lr_mult: 2.0
decay_mult: 0.0
}
convolution_param {
num_output:
pad:
kernel_size:
stride:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "fc7_mbox_conf_perm"
type: "Permute"
bottom: "fc7_mbox_conf"
top: "fc7_mbox_conf_perm"
permute_param {
order:
order:
order:
order:
}
}
layer {
name: "fc7_mbox_conf_flat"
type: "Flatten"
bottom: "fc7_mbox_conf_perm"
top: "fc7_mbox_conf_flat"
flatten_param {
axis:
}
}
layer {
name: "fc7_mbox_priorbox"
type: "PriorBox"
bottom: "fc7"
bottom: "data"
top: "fc7_mbox_priorbox"
prior_box_param {
min_size: 64.0
aspect_ratio: 2.0
flip: true
clip: false
variance: 0.10000000149
variance: 0.10000000149
variance: 0.20000000298
variance: 0.20000000298
step: 32.0
offset: 0.5
}
}
layer {
name: "conv6_2_mbox_loc"
type: "Convolution"
bottom: "conv6_2"
top: "conv6_2_mbox_loc"
param {
lr_mult: 1.0
decay_mult: 1.0
}
param {
lr_mult: 2.0
decay_mult: 0.0
}
convolution_param {
num_output:
pad:
kernel_size:
stride:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "conv6_2_mbox_loc_perm"
type: "Permute"
bottom: "conv6_2_mbox_loc"
top: "conv6_2_mbox_loc_perm"
permute_param {
order:
order:
order:
order:
}
}
layer {
name: "conv6_2_mbox_loc_flat"
type: "Flatten"
bottom: "conv6_2_mbox_loc_perm"
top: "conv6_2_mbox_loc_flat"
flatten_param {
axis:
}
}
layer {
name: "conv6_2_mbox_conf"
type: "Convolution"
bottom: "conv6_2"
top: "conv6_2_mbox_conf"
param {
lr_mult: 1.0
decay_mult: 1.0
}
param {
lr_mult: 2.0
decay_mult: 0.0
}
convolution_param {
num_output:
pad:
kernel_size:
stride:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "conv6_2_mbox_conf_perm"
type: "Permute"
bottom: "conv6_2_mbox_conf"
top: "conv6_2_mbox_conf_perm"
permute_param {
order:
order:
order:
order:
}
}
layer {
name: "conv6_2_mbox_conf_flat"
type: "Flatten"
bottom: "conv6_2_mbox_conf_perm"
top: "conv6_2_mbox_conf_flat"
flatten_param {
axis:
}
}
layer {
name: "conv6_2_mbox_priorbox"
type: "PriorBox"
bottom: "conv6_2"
bottom: "data"
top: "conv6_2_mbox_priorbox"
prior_box_param {
min_size: 128.0
aspect_ratio: 2.0
flip: true
clip: false
variance: 0.10000000149
variance: 0.10000000149
variance: 0.20000000298
variance: 0.20000000298
step: 64.0
offset: 0.5
}
}
layer {
name: "arm_loc"
type: "Concat"
bottom: "conv4_3_norm_mbox_loc_flat"
bottom: "conv5_3_norm_mbox_loc_flat"
bottom: "fc7_mbox_loc_flat"
bottom: "conv6_2_mbox_loc_flat"
top: "arm_loc"
concat_param {
axis:
}
}
layer {
name: "arm_conf"
type: "Concat"
bottom: "conv4_3_norm_mbox_conf_flat"
bottom: "conv5_3_norm_mbox_conf_flat"
bottom: "fc7_mbox_conf_flat"
bottom: "conv6_2_mbox_conf_flat"
top: "arm_conf"
concat_param {
axis:
}
}
layer {
name: "arm_priorbox"
type: "Concat"
bottom: "conv4_3_norm_mbox_priorbox"
bottom: "conv5_3_norm_mbox_priorbox"
bottom: "fc7_mbox_priorbox"
bottom: "conv6_2_mbox_priorbox"
top: "arm_priorbox"
concat_param {
axis:
}
}
layer {
name: "P3_mbox_loc_p"
type: "Convolution"
bottom: "conv4_3"
top: "P3_mbox_loc"
param {
lr_mult: 1.0
decay_mult: 1.0
}
param {
lr_mult: 2.0
decay_mult: 0.0
}
convolution_param {
num_output:
pad:
kernel_size:
stride:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "P3_mbox_loc_perm"
type: "Permute"
bottom: "P3_mbox_loc"
top: "P3_mbox_loc_perm"
permute_param {
order:
order:
order:
order:
}
}
layer {
name: "P3_mbox_loc_flat"
type: "Flatten"
bottom: "P3_mbox_loc_perm"
top: "P3_mbox_loc_flat"
flatten_param {
axis:
}
}
layer {
name: "P3_mbox_conf_p"
type: "Convolution"
bottom: "conv4_3"
top: "P3_mbox_conf"
param {
lr_mult: 1.0
decay_mult: 1.0
}
param {
lr_mult: 2.0
decay_mult: 0.0
}
convolution_param {
num_output:
pad:
kernel_size:
stride:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "P3_mbox_conf_perm"
type: "Permute"
bottom: "P3_mbox_conf"
top: "P3_mbox_conf_perm"
permute_param {
order:
order:
order:
order:
}
}
layer {
name: "P3_mbox_conf_flat"
type: "Flatten"
bottom: "P3_mbox_conf_perm"
top: "P3_mbox_conf_flat"
flatten_param {
axis:
}
}
layer {
name: "P4_mbox_loc_p"
type: "Convolution"
bottom: "conv5_3"
top: "P4_mbox_loc"
param {
lr_mult: 1.0
decay_mult: 1.0
}
param {
lr_mult: 2.0
decay_mult: 0.0
}
convolution_param {
num_output:
pad:
kernel_size:
stride:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "P4_mbox_loc_perm"
type: "Permute"
bottom: "P4_mbox_loc"
top: "P4_mbox_loc_perm"
permute_param {
order:
order:
order:
order:
}
}
layer {
name: "P4_mbox_loc_flat"
type: "Flatten"
bottom: "P4_mbox_loc_perm"
top: "P4_mbox_loc_flat"
flatten_param {
axis:
}
}
layer {
name: "P4_mbox_conf_p"
type: "Convolution"
bottom: "conv5_3"
top: "P4_mbox_conf"
param {
lr_mult: 1.0
decay_mult: 1.0
}
param {
lr_mult: 2.0
decay_mult: 0.0
}
convolution_param {
num_output:
pad:
kernel_size:
stride:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "P4_mbox_conf_perm"
type: "Permute"
bottom: "P4_mbox_conf"
top: "P4_mbox_conf_perm"
permute_param {
order:
order:
order:
order:
}
}
layer {
name: "P4_mbox_conf_flat"
type: "Flatten"
bottom: "P4_mbox_conf_perm"
top: "P4_mbox_conf_flat"
flatten_param {
axis:
}
}
layer {
name: "P5_mbox_loc_p"
type: "Convolution"
bottom: "fc7"
top: "P5_mbox_loc"
param {
lr_mult: 1.0
decay_mult: 1.0
}
param {
lr_mult: 2.0
decay_mult: 0.0
}
convolution_param {
num_output:
pad:
kernel_size:
stride:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "P5_mbox_loc_perm"
type: "Permute"
bottom: "P5_mbox_loc"
top: "P5_mbox_loc_perm"
permute_param {
order:
order:
order:
order:
}
}
layer {
name: "P5_mbox_loc_flat"
type: "Flatten"
bottom: "P5_mbox_loc_perm"
top: "P5_mbox_loc_flat"
flatten_param {
axis:
}
}
layer {
name: "P5_mbox_conf_p"
type: "Convolution"
bottom: "fc7"
top: "P5_mbox_conf"
param {
lr_mult: 1.0
decay_mult: 1.0
}
param {
lr_mult: 2.0
decay_mult: 0.0
}
convolution_param {
num_output:
pad:
kernel_size:
stride:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "P5_mbox_conf_perm"
type: "Permute"
bottom: "P5_mbox_conf"
top: "P5_mbox_conf_perm"
permute_param {
order:
order:
order:
order:
}
}
layer {
name: "P5_mbox_conf_flat"
type: "Flatten"
bottom: "P5_mbox_conf_perm"
top: "P5_mbox_conf_flat"
flatten_param {
axis:
}
}
layer {
name: "P6_mbox_loc_p"
type: "Convolution"
bottom: "conv6_2"
top: "P6_mbox_loc"
param {
lr_mult: 1.0
decay_mult: 1.0
}
param {
lr_mult: 2.0
decay_mult: 0.0
}
convolution_param {
num_output:
pad:
kernel_size:
stride:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "P6_mbox_loc_perm"
type: "Permute"
bottom: "P6_mbox_loc"
top: "P6_mbox_loc_perm"
permute_param {
order:
order:
order:
order:
}
}
layer {
name: "P6_mbox_loc_flat"
type: "Flatten"
bottom: "P6_mbox_loc_perm"
top: "P6_mbox_loc_flat"
flatten_param {
axis:
}
}
layer {
name: "P6_mbox_conf_p"
type: "Convolution"
bottom: "conv6_2"
top: "P6_mbox_conf"
param {
lr_mult: 1.0
decay_mult: 1.0
}
param {
lr_mult: 2.0
decay_mult: 0.0
}
convolution_param {
num_output:
pad:
kernel_size:
stride:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "P6_mbox_conf_perm"
type: "Permute"
bottom: "P6_mbox_conf"
top: "P6_mbox_conf_perm"
permute_param {
order:
order:
order:
order:
}
}
layer {
name: "P6_mbox_conf_flat"
type: "Flatten"
bottom: "P6_mbox_conf_perm"
top: "P6_mbox_conf_flat"
flatten_param {
axis:
}
}
layer {
name: "odm_loc"
type: "Concat"
bottom: "P3_mbox_loc_flat"
bottom: "P4_mbox_loc_flat"
bottom: "P5_mbox_loc_flat"
bottom: "P6_mbox_loc_flat"
top: "odm_loc"
concat_param {
axis:
}
}
layer {
name: "odm_conf"
type: "Concat"
bottom: "P3_mbox_conf_flat"
bottom: "P4_mbox_conf_flat"
bottom: "P5_mbox_conf_flat"
bottom: "P6_mbox_conf_flat"
top: "odm_conf"
concat_param {
axis:
}
}
layer {
name: "arm_loss"
type: "MultiBoxLoss"
bottom: "arm_loc"
bottom: "arm_conf"
bottom: "arm_priorbox"
bottom: "label"
top: "arm_loss"
include {
phase: TRAIN
}
propagate_down: true
propagate_down: true
propagate_down: false
propagate_down: false
loss_param {
normalization: VALID
}
multibox_loss_param {
loc_loss_type: SMOOTH_L1
conf_loss_type: SOFTMAX
loc_weight: 1.0
num_classes:
share_location: true
match_type: PER_PREDICTION
overlap_threshold: 0.5
use_prior_for_matching: true
background_label_id:
use_difficult_gt: true
neg_pos_ratio: 3.0
neg_overlap: 0.5
code_type: CENTER_SIZE
ignore_cross_boundary_bbox: false
mining_type: MAX_NEGATIVE
objectness_score: 0.00999999977648
}
}
layer {
name: "arm_conf_reshape"
type: "Reshape"
bottom: "arm_conf"
top: "arm_conf_reshape"
reshape_param {
shape {
dim:
dim: -
dim:
}
}
}
layer {
name: "arm_conf_softmax"
type: "Softmax"
bottom: "arm_conf_reshape"
top: "arm_conf_softmax"
softmax_param {
axis:
}
}
layer {
name: "arm_conf_flatten"
type: "Flatten"
bottom: "arm_conf_softmax"
top: "arm_conf_flatten"
flatten_param {
axis:
}
}
layer {
name: "odm_loss"
type: "MultiBoxLoss"
bottom: "odm_loc"
bottom: "odm_conf"
bottom: "arm_priorbox"
bottom: "label"
bottom: "arm_conf_flatten"
bottom: "arm_loc"
top: "odm_loss"
include {
phase: TRAIN
}
propagate_down: true
propagate_down: true
propagate_down: false
propagate_down: false
propagate_down: false
propagate_down: false
loss_param {
normalization: VALID
}
multibox_loss_param {
loc_loss_type: SMOOTH_L1
conf_loss_type: SOFTMAX
loc_weight: 1.0
num_classes:
share_location: true
match_type: PER_PREDICTION
overlap_threshold: 0.5
use_prior_for_matching: true
background_label_id:
use_difficult_gt: true
neg_pos_ratio: 3.0
neg_overlap: 0.5
code_type: CENTER_SIZE
ignore_cross_boundary_bbox: false
mining_type: MAX_NEGATIVE
objectness_score: 0.00999999977648
}
} layer {
name: "conv1_1_t"
type: "Convolution"
bottom: "data"
top: "conv1_1_t"
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
convolution_param {
num_output:
pad:
kernel_size:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "relu1_1_t"
type: "ReLU"
bottom: "conv1_1_t"
top: "conv1_1_t"
}
layer {
name: "conv1_2_t"
type: "Convolution"
bottom: "conv1_1_t"
top: "conv1_2_t"
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
convolution_param {
num_output:
pad:
kernel_size:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "relu1_2_t"
type: "ReLU"
bottom: "conv1_2_t"
top: "conv1_2_t"
}
layer {
name: "pool1_t"
type: "Pooling"
bottom: "conv1_2_t"
top: "pool1_t"
pooling_param {
pool: MAX
kernel_size:
stride:
}
}
layer {
name: "conv2_1_t"
type: "Convolution"
bottom: "pool1_t"
top: "conv2_1_t"
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
convolution_param {
num_output:
pad:
kernel_size:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "relu2_1_t"
type: "ReLU"
bottom: "conv2_1_t"
top: "conv2_1_t"
}
layer {
name: "conv2_2_t"
type: "Convolution"
bottom: "conv2_1_t"
top: "conv2_2_t"
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
convolution_param {
num_output:
pad:
kernel_size:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "relu2_2_t"
type: "ReLU"
bottom: "conv2_2_t"
top: "conv2_2_t"
}
layer {
name: "pool2_t"
type: "Pooling"
bottom: "conv2_2_t"
top: "pool2_t"
pooling_param {
pool: MAX
kernel_size:
stride:
}
}
layer {
name: "conv3_1_t"
type: "Convolution"
bottom: "pool2_t"
top: "conv3_1_t"
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
convolution_param {
num_output:
pad:
kernel_size:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "relu3_1_t"
type: "ReLU"
bottom: "conv3_1_t"
top: "conv3_1_t"
}
layer {
name: "conv3_2_t"
type: "Convolution"
bottom: "conv3_1_t"
top: "conv3_2_t"
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
convolution_param {
num_output:
pad:
kernel_size:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "relu3_2_t"
type: "ReLU"
bottom: "conv3_2_t"
top: "conv3_2_t"
}
layer {
name: "conv3_3_t"
type: "Convolution"
bottom: "conv3_2_t"
top: "conv3_3_t"
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
convolution_param {
num_output:
pad:
kernel_size:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "relu3_3_t"
type: "ReLU"
bottom: "conv3_3_t"
top: "conv3_3_t"
}
layer {
name: "pool3_t"
type: "Pooling"
bottom: "conv3_3_t"
top: "pool3_t"
pooling_param {
pool: MAX
kernel_size:
stride:
}
}
layer {
name: "conv4_1_t"
type: "Convolution"
bottom: "pool3_t"
top: "conv4_1_t"
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
convolution_param {
num_output:
pad:
kernel_size:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "relu4_1_t"
type: "ReLU"
bottom: "conv4_1_t"
top: "conv4_1_t"
}
layer {
name: "conv4_2_t"
type: "Convolution"
bottom: "conv4_1_t"
top: "conv4_2_t"
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
convolution_param {
num_output:
pad:
kernel_size:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "relu4_2_t"
type: "ReLU"
bottom: "conv4_2_t"
top: "conv4_2_t"
}
layer {
name: "conv4_3_t"
type: "Convolution"
bottom: "conv4_2_t"
top: "conv4_3_t"
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
convolution_param {
num_output:
pad:
kernel_size:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
}
}
layer {
name: "relu4_3_t"
type: "ReLU"
bottom: "conv4_3_t"
top: "conv4_3_t"
}
layer {
name: "pool4_t"
type: "Pooling"
bottom: "conv4_3_t"
top: "pool4_t"
pooling_param {
pool: MAX
kernel_size:
stride:
}
}
layer {
name: "conv5_1_t"
type: "Convolution"
bottom: "pool4_t"
top: "conv5_1_t"
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
convolution_param {
num_output:
pad:
kernel_size:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
dilation:
}
}
layer {
name: "relu5_1_t"
type: "ReLU"
bottom: "conv5_1_t"
top: "conv5_1_t"
}
layer {
name: "conv5_2_t"
type: "Convolution"
bottom: "conv5_1_t"
top: "conv5_2_t"
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
convolution_param {
num_output:
pad:
kernel_size:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
dilation:
}
}
layer {
name: "relu5_2_t"
type: "ReLU"
bottom: "conv5_2_t"
top: "conv5_2_t"
}
layer {
name: "conv5_3_t"
type: "Convolution"
bottom: "conv5_2_t"
top: "conv5_3_t"
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
convolution_param {
num_output:
pad:
kernel_size:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value: 0.0
}
dilation:
}
}
layer {
name: "relu5_3_t"
type: "ReLU"
bottom: "conv5_3_t"
top: "conv5_3_t"
} layer {
name: "conv5_3_m"
type: "Convolution"
bottom: "conv5_3"
top: "conv5_3_m"
propagate_down: true
param {
lr_mult:
decay_mult:
}
param {
lr_mult:
decay_mult:
}
convolution_param {
num_output:
kernel_size:
weight_filler {
type: "xavier"
}
bias_filler {
type: "constant"
value:
}
}
} layer {
name: "relu5_3_m"
type: "ReLU"
bottom: "conv5_3_m"
top: "conv5_3_m"
} layer {
name: "roi_pool_t"
type: "ROIPooling"
bottom: "conv5_3_t"
bottom: "label"
top: "pool_t"
roi_pooling_param {
pooled_w:
pooled_h:
}
propagate_down: false
propagate_down: false
}
layer {
name: "roi_pool_s"
type: "ROIPooling"
bottom: "conv5_3_m"
bottom: "label"
top: "pool_s"
roi_pooling_param {
pooled_w:
pooled_h:
}
propagate_down: true
propagate_down: false
} layer {
name: "mimic_loss"
type: "EuclideanLoss"
bottom: "pool_t"
bottom: "pool_s"
top: "mimic_loss"
propagate_down: false
propagate_down: true
loss_weight:
include {
phase: TRAIN
}
}
refinedet网络结构的更多相关文章
- RefineDet算法笔记
---恢复内容开始--- 一.创新点 针对two-stage的速度慢以及one-stage精度不足提出的方法,refinedet 包括三个核心部分:使用TCB来转换ARM的特征,送入ODM中进行检测: ...
- 论文阅读 | RefineDet：Single-Shot Refinement Neural Network for Object Detection
论文链接:https://arxiv.org/abs/1711.06897 代码链接:https://github.com/sfzhang15/RefineDet 摘要 RefineDet是CVPR ...
- 目标检测之RefineDet
RefineDet 一.相关背景 中科院自动化所最新成果,CVPR 2018 <Single-Shot Refinement Neural Network for Object Detectio ...
- Sparse Filtering 学习笔记（一）网络结构与特征矩阵
Sparse Filtering 是一个用于提取特征的无监督学习算法,与通常特征学习算法试图建模训练数据的分布的做法不同,Sparse Filtering 直接对训练数据的特征分布进行分析,在所谓 ...
- 深度神经网络结构以及Pre-Training的理解
Logistic回归.传统多层神经网络 1.1 线性回归.线性神经网络.Logistic/Softmax回归 线性回归是用于数据拟合的常规手段,其任务是优化目标函数:$h(\theta )=\thet ...
- 受限玻尔兹曼机（RBM）学习笔记（二）网络结构
去年 6 月份写的博文<Yusuke Sugomori 的 C 语言 Deep Learning 程序解读>是囫囵吞枣地读完一个关于 DBN 算法的开源代码后的笔记,当时对其中涉及的算 ...
- OpenWrt网络结构
原文链接:http://www.freezhongzi.info/?p=104 OpenWrt网络结构 OpenWrt的网络配置很丰富,在我看来几乎可以完成任何网络结构.下图为一个支持OpenWrt的 ...
- CNN(卷积神经网络)、RNN(循环神经网络)、DNN(深度神经网络)的内部网络结构有什么区别？
https://www.zhihu.com/question/34681168 CNN(卷积神经网络).RNN(循环神经网络).DNN(深度神经网络)的内部网络结构有什么区别?修改 CNN(卷积神经网 ...
- 神经网络结构在命名实体识别（NER）中的应用
神经网络结构在命名实体识别(NER)中的应用 近年来,基于神经网络的深度学习方法在自然语言处理领域已经取得了不少进展.作为NLP领域的基础任务-命名实体识别(Named Entity Recognit ...
随机推荐
- 在Spark shell中基于HDFS文件系统进行wordcount交互式分析
Spark是一个分布式内存计算框架,可部署在YARN或者MESOS管理的分布式系统中(Fully Distributed),也可以以Pseudo Distributed方式部署在单个机器上面,还可以以 ...
- 吴恩达《Machine Learning Yearning》总结（1-10章）
1.为什么选择机器学习策略 案例:建立猫咪图像识别app 系统的优化可以有很多的方向: (1)获取更多的数据集,即更多的图片: (2)收集更多多样数据,如处于不常见的位置的猫的图,颜色奇异的猫的照片等 ...
- Javascript模块化编程(二)AMD规范(规范使用模块)
这个系列的第一部分介绍了Javascript模块的基本写法,今天介绍如何规范地使用模块,先想一想,为什么模块很重要?接下来为您详细介绍,感兴趣的朋友可以了解下啊.今天介绍如何规范地使用模块. 七.模块 ...
- 使用ANY、Some或All关键字
可以使用All或Any关键字修改引入子查询的比较运算符.Some是与Any等效的ISO标准,All要求Where表达式与子查询返回的每个值进行比较时都应满足比较条件,Any则要求Where表达式与子查 ...
- django基本入门
1.创建应用 2.设计模型 3.语言时区等设置 4. Templates 1.创建应用[MVT] 一个项目可以有多个应用[模块]: 这里已经创建了项目:test1 python manager.py ...
- js把数据处理成钱的格式
1.var Rmoney = parseFloat(money).toFixed(2);//把money处理成保存2位小数的格式
- 重构指南 - 引入参数对象（Introduce Parameter Object）
当一个方法的参数超过3个以上,就可以考虑将参数封装成一个对象.将参数封装成对象后提高了代码的可读性,并且该参数对象也可以供多个方法调用,以后如果增加删除参数,方法本身不需要修改,只需要修改参数对象就可 ...
- PAT 1064 Complete Binary Search Tree
#include <iostream> #include <cstdio> #include <cstdlib> #include <vector> # ...
- PHP连接MySQL数据库的几种方式
PHP 5 及以上版本建议使用以下方式连接 MySQL : MySQLi :MySQLi 只针对 MySQL 数据库,MySQLi 还提供了 API 接口. PDO (PHP Data Objects ...
- Flexviewer使用Google地图作为底图
Flexviewer使用Google地图作为底图: 在使用google地图作底图前提是你需要在Flex中实现加载google地图的代码(网上一大堆,随便找), 在只加载google地图的情况下,成功显 ...