Last active
April 7, 2018 12:41
-
-
Save jens25/6b0ea1143599fb99bd499a08dd5c072c to your computer and use it in GitHub Desktop.
Trainining and validation files for the RACNN Network.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
name: "RA_CNN" | |
layer { | |
name: "data" | |
type: "Data" | |
top: "data" | |
top: "label" | |
include { | |
phase: TRAIN | |
} | |
transform_param { | |
mirror: true | |
crop_size: 448 | |
mean_value: 128 | |
mean_value: 128 | |
mean_value: 128 | |
} | |
data_param { | |
source: "/media/data/lmdb/birds" | |
batch_size: 2 | |
backend: LMDB | |
} | |
} | |
layer { | |
name: "data" | |
type: "Data" | |
top: "data" | |
top: "label" | |
include { | |
phase: TEST | |
} | |
transform_param { | |
mirror: true | |
crop_size: 448 | |
mean_value: 128 | |
mean_value: 128 | |
mean_value: 128 | |
} | |
data_param { | |
source: "/media/data/lmdb/birds" | |
batch_size: 2 | |
backend: LMDB | |
} | |
} | |
#######Scale1####### | |
layer { | |
bottom: "data" | |
top: "conv1_1" | |
name: "conv1_1" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 64 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv1_1" | |
top: "conv1_1" | |
name: "relu1_1" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv1_1" | |
top: "conv1_2" | |
name: "conv1_2" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 64 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv1_2" | |
top: "conv1_2" | |
name: "relu1_2" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv1_2" | |
top: "pool1" | |
name: "pool1" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool1" | |
top: "conv2_1" | |
name: "conv2_1" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 128 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv2_1" | |
top: "conv2_1" | |
name: "relu2_1" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv2_1" | |
top: "conv2_2" | |
name: "conv2_2" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 128 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv2_2" | |
top: "conv2_2" | |
name: "relu2_2" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv2_2" | |
top: "pool2" | |
name: "pool2" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool2" | |
top: "conv3_1" | |
name: "conv3_1" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_1" | |
top: "conv3_1" | |
name: "relu3_1" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_1" | |
top: "conv3_2" | |
name: "conv3_2" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_2" | |
top: "conv3_2" | |
name: "relu3_2" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_2" | |
top: "conv3_3" | |
name: "conv3_3" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_3" | |
top: "conv3_3" | |
name: "relu3_3" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_3" | |
top: "conv3_4" | |
name: "conv3_4" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_4" | |
top: "conv3_4" | |
name: "relu3_4" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_4" | |
top: "pool3" | |
name: "pool3" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool3" | |
top: "conv4_1" | |
name: "conv4_1" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_1" | |
top: "conv4_1" | |
name: "relu4_1" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_1" | |
top: "conv4_2" | |
name: "conv4_2" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_2" | |
top: "conv4_2" | |
name: "relu4_2" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_2" | |
top: "conv4_3" | |
name: "conv4_3" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_3" | |
top: "conv4_3" | |
name: "relu4_3" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_3" | |
top: "conv4_4" | |
name: "conv4_4" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_4" | |
top: "conv4_4" | |
name: "relu4_4" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_4" | |
top: "pool4" | |
name: "pool4" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool4" | |
top: "conv5_1" | |
name: "conv5_1" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_1" | |
top: "conv5_1" | |
name: "relu5_1" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv5_1" | |
top: "conv5_2" | |
name: "conv5_2" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_2" | |
top: "conv5_2" | |
name: "relu5_2" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv5_2" | |
top: "conv5_3" | |
name: "conv5_3" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_3" | |
top: "conv5_3" | |
name: "relu5_3" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv5_3" | |
top: "conv5_4" | |
name: "conv5_4" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_4" | |
top: "conv5_4" | |
name: "relu5_4" | |
type: "ReLU" | |
} | |
layer { | |
name: "pool5" | |
type: "Pooling" | |
bottom: "conv5_4" | |
top: "pool5" | |
pooling_param { | |
pool: AVE | |
kernel_size: 28 | |
stride: 28 | |
} | |
} | |
#######APN1####### | |
layer { | |
bottom: "conv5_4" | |
top: "anp_pool" | |
name: "anp_pool" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
name: "get_abc1" | |
type: "InnerProduct" | |
bottom: "anp_pool" | |
top: "get_abc1" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
inner_product_param { | |
num_output: 1024 | |
weight_filler { | |
type: "gaussian" | |
std: 0.001 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
name: "tanh" | |
bottom: "get_abc1" | |
top: "tanh" | |
type: "TanH" | |
} | |
layer { | |
name: "get_abc2" | |
type: "InnerProduct" | |
bottom: "tanh" | |
top: "get_abc2" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
inner_product_param { | |
num_output: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.001 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
name: "sigmoid" | |
bottom: "get_abc2" | |
top: "sig_abc" | |
type: "Sigmoid" | |
} | |
#######Scale2####### | |
layer { | |
name: "get448" | |
bottom: "sig_abc" | |
top: "get448" | |
type: "Power" | |
power_param { | |
power: 1 | |
scale: 448 | |
shift: 0 | |
} | |
} | |
layer{ | |
name: "atten_crop" | |
bottom: "data" | |
bottom: "get448" | |
top: "scale2_data" | |
type: "AttentionCrop" | |
} | |
layer { | |
bottom: "scale2_data" | |
top: "conv1_1_A" | |
name: "conv1_1_A" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 64 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv1_1_A" | |
top: "conv1_1_A" | |
name: "relu1_1_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv1_1_A" | |
top: "conv1_2_A" | |
name: "conv1_2_A" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 64 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv1_2_A" | |
top: "conv1_2_A" | |
name: "relu1_2_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv1_2_A" | |
top: "pool1_A" | |
name: "pool1_A" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool1_A" | |
top: "conv2_1_A" | |
name: "conv2_1_A" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 128 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv2_1_A" | |
top: "conv2_1_A" | |
name: "relu2_1_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv2_1_A" | |
top: "conv2_2_A" | |
name: "conv2_2_A" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 128 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv2_2_A" | |
top: "conv2_2_A" | |
name: "relu2_2_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv2_2_A" | |
top: "pool2_A" | |
name: "pool2_A" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool2_A" | |
top: "conv3_1_A" | |
name: "conv3_1_A" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_1_A" | |
top: "conv3_1_A" | |
name: "relu3_1_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_1_A" | |
top: "conv3_2_A" | |
name: "conv3_2_A" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_2_A" | |
top: "conv3_2_A" | |
name: "relu3_2_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_2_A" | |
top: "conv3_3_A" | |
name: "conv3_3_A" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_3_A" | |
top: "conv3_3_A" | |
name: "relu3_3_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_3_A" | |
top: "conv3_4_A" | |
name: "conv3_4_A" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_4_A" | |
top: "conv3_4_A" | |
name: "relu3_4_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_4_A" | |
top: "pool3_A" | |
name: "pool3_A" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool3_A" | |
top: "conv4_1_A" | |
name: "conv4_1_A" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_1_A" | |
top: "conv4_1_A" | |
name: "relu4_1_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_1_A" | |
top: "conv4_2_A" | |
name: "conv4_2_A" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_2_A" | |
top: "conv4_2_A" | |
name: "relu4_2_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_2_A" | |
top: "conv4_3_A" | |
name: "conv4_3_A" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_3_A" | |
top: "conv4_3_A" | |
name: "relu4_3_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_3_A" | |
top: "conv4_4_A" | |
name: "conv4_4_A" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_4_A" | |
top: "conv4_4_A" | |
name: "relu4_4_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_4_A" | |
top: "pool4_A" | |
name: "pool4_A" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool4_A" | |
top: "conv5_1_A" | |
name: "conv5_1_A" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_1_A" | |
top: "conv5_1_A" | |
name: "relu5_1_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv5_1_A" | |
top: "conv5_2_A" | |
name: "conv5_2_A" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_2_A" | |
top: "conv5_2_A" | |
name: "relu5_2_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv5_2_A" | |
top: "conv5_3_A" | |
name: "conv5_3_A" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_3_A" | |
top: "conv5_3_A" | |
name: "relu5_3_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv5_3_A" | |
top: "conv5_4_A" | |
name: "conv5_4_A" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_4_A" | |
top: "conv5_4_A" | |
name: "relu5_4_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv5_4_A" | |
top: "pool5_A" | |
name: "pool5_A" | |
type: "Pooling" | |
pooling_param { | |
pool: AVE | |
kernel_size: 14 | |
stride: 14 | |
} | |
} | |
#######APN2####### | |
layer { | |
name: "get_abc1_A" | |
type: "InnerProduct" | |
bottom: "conv5_4_A" | |
top: "get_abc1_A" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
inner_product_param { | |
num_output: 1024 | |
weight_filler { | |
type: "gaussian" | |
std: 0.001 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
name: "tanh_A" | |
bottom: "get_abc1_A" | |
top: "tanh_A" | |
type: "TanH" | |
} | |
layer { | |
name: "get_abc2_A" | |
type: "InnerProduct" | |
bottom: "tanh_A" | |
top: "get_abc2_A" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
inner_product_param { | |
num_output: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.001 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
name: "sigmoid_A" | |
bottom: "get_abc2_A" | |
top: "sig_abc_A" | |
type: "Sigmoid" | |
} | |
#######Scale3####### | |
layer { | |
name: "get224" | |
bottom: "sig_abc_A" | |
top: "get224" | |
type: "Power" | |
power_param { | |
power: 1 | |
scale: 224 | |
shift: 0 | |
} | |
} | |
layer{ | |
name: "atten_crop_A" | |
bottom: "scale2_data" | |
bottom: "get224" | |
top: "scale3_data" | |
type: "AttentionCrop" | |
} | |
layer { | |
bottom: "scale3_data" | |
top: "conv1_1_A_A" | |
name: "conv1_1_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 64 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv1_1_A_A" | |
top: "conv1_1_A_A" | |
name: "relu1_1_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv1_1_A_A" | |
top: "conv1_2_A_A" | |
name: "conv1_2_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 64 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv1_2_A_A" | |
top: "conv1_2_A_A" | |
name: "relu1_2_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv1_2_A_A" | |
top: "pool1_A_A" | |
name: "pool1_A_A" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool1_A_A" | |
top: "conv2_1_A_A" | |
name: "conv2_1_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 128 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv2_1_A_A" | |
top: "conv2_1_A_A" | |
name: "relu2_1_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv2_1_A_A" | |
top: "conv2_2_A_A" | |
name: "conv2_2_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 128 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv2_2_A_A" | |
top: "conv2_2_A_A" | |
name: "relu2_2_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv2_2_A_A" | |
top: "pool2_A_A" | |
name: "pool2_A_A" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool2_A_A" | |
top: "conv3_1_A_A" | |
name: "conv3_1_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_1_A_A" | |
top: "conv3_1_A_A" | |
name: "relu3_1_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_1_A_A" | |
top: "conv3_2_A_A" | |
name: "conv3_2_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_2_A_A" | |
top: "conv3_2_A_A" | |
name: "relu3_2_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_2_A_A" | |
top: "conv3_3_A_A" | |
name: "conv3_3_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_3_A_A" | |
top: "conv3_3_A_A" | |
name: "relu3_3_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_3_A_A" | |
top: "conv3_4_A_A" | |
name: "conv3_4_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_4_A_A" | |
top: "conv3_4_A_A" | |
name: "relu3_4_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_4_A_A" | |
top: "pool3_A_A" | |
name: "pool3_A_A" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool3_A_A" | |
top: "conv4_1_A_A" | |
name: "conv4_1_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_1_A_A" | |
top: "conv4_1_A_A" | |
name: "relu4_1_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_1_A_A" | |
top: "conv4_2_A_A" | |
name: "conv4_2_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_2_A_A" | |
top: "conv4_2_A_A" | |
name: "relu4_2_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_2_A_A" | |
top: "conv4_3_A_A" | |
name: "conv4_3_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_3_A_A" | |
top: "conv4_3_A_A" | |
name: "relu4_3_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_3_A_A" | |
top: "conv4_4_A_A" | |
name: "conv4_4_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_4_A_A" | |
top: "conv4_4_A_A" | |
name: "relu4_4_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_4_A_A" | |
top: "pool4_A_A" | |
name: "pool4_A_A" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool4_A_A" | |
top: "conv5_1_A_A" | |
name: "conv5_1_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_1_A_A" | |
top: "conv5_1_A_A" | |
name: "relu5_1_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv5_1_A_A" | |
top: "conv5_2_A_A" | |
name: "conv5_2_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_2_A_A" | |
top: "conv5_2_A_A" | |
name: "relu5_2_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv5_2_A_A" | |
top: "conv5_3_A_A" | |
name: "conv5_3_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_3_A_A" | |
top: "conv5_3_A_A" | |
name: "relu5_3_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv5_3_A_A" | |
top: "conv5_4_A_A" | |
name: "conv5_4_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_4_A_A" | |
top: "conv5_4_A_A" | |
name: "relu5_4_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv5_4_A_A" | |
top: "pool5_A_A" | |
name: "pool5_A_A" | |
type: "Pooling" | |
pooling_param { | |
pool: AVE | |
kernel_size: 14 | |
stride: 14 | |
} | |
} | |
#####feature_fusion##### | |
layer { | |
name: "reshape1" | |
bottom: "pool5" | |
top: "reshape1" | |
type: "Reshape" | |
reshape_param { | |
shape { | |
dim: -1 | |
dim: 512 | |
} | |
} | |
} | |
layer { | |
name: "reshape2" | |
bottom: "pool5_A" | |
top: "reshape2" | |
type: "Reshape" | |
reshape_param { | |
shape { | |
dim: -1 | |
dim: 512 | |
} | |
} | |
} | |
layer { | |
name: "reshape3" | |
bottom: "pool5_A_A" | |
top: "reshape3" | |
type: "Reshape" | |
reshape_param { | |
shape { | |
dim: -1 | |
dim: 512 | |
} | |
} | |
} | |
layer { | |
name: "pow1" | |
bottom: "reshape1" | |
top: "pow1" | |
type: "Power" | |
power_param { | |
power: 1 | |
scale: 0.1 | |
shift: 0 | |
} | |
} | |
layer { | |
name: "pow2" | |
bottom: "reshape2" | |
top: "pow2" | |
type: "Power" | |
power_param { | |
power: 1 | |
scale: 0.1 | |
shift: 0 | |
} | |
} | |
layer { | |
name: "pow3" | |
bottom: "reshape3" | |
top: "pow3" | |
type: "Power" | |
power_param { | |
power: 1 | |
scale: 0.1 | |
shift: 0 | |
} | |
} | |
#layer { | |
# name: "scale1+2+3" | |
# bottom: "pow2" | |
# bottom: "pow1" | |
# bottom: "pow3" | |
# top: "scale1+2+3" | |
# type: "Concat" | |
# concat_param { | |
# axis: 1 | |
# } | |
#} | |
#layer { | |
# name: "scale1+2" | |
# bottom: "pow2" | |
# bottom: "pow1" | |
# top: "scale1+2" | |
# type: "Concat" | |
# concat_param { | |
# axis: 1 | |
# } | |
#} | |
#layer { | |
# name: "fc1_custom" | |
# type: "InnerProduct" | |
# bottom: "scale1+2+3" | |
# top: "fc1_custom" | |
# param { | |
# lr_mult: 1.0 | |
# decay_mult: 0 | |
# } | |
# param { | |
# lr_mult: 1.0 | |
# decay_mult: 0 | |
# } | |
# inner_product_param { | |
# num_output: 100 | |
# weight_filler { | |
# type: "gaussian" | |
# std: 0.01 | |
# } | |
# bias_filler { | |
# type: "constant" | |
# value: 0 | |
# } | |
# } | |
#} | |
#layer { | |
# name: "accuracy1+2+3" | |
# type: "Accuracy" | |
# bottom: "fc1_custom" | |
# bottom: "label" | |
# top: "accuracy1+2+3" | |
# include { | |
# phase: TEST | |
# } | |
#} | |
#layer { | |
# name: "fc2_custom" | |
# type: "InnerProduct" | |
# bottom: "scale1+2" | |
# top: "fc2_custom" | |
# param { | |
# lr_mult: 1.0 | |
# decay_mult: 0 | |
# } | |
# param { | |
# lr_mult: 1.0 | |
# decay_mult: 0 | |
# } | |
# inner_product_param { | |
# num_output: 100 | |
# weight_filler { | |
# type: "gaussian" | |
# std: 0.01 | |
# } | |
# bias_filler { | |
# type: "constant" | |
# value: 0 | |
# } | |
# } | |
#} | |
#layer { | |
# name: "accuracy1+2" | |
# type: "Accuracy" | |
# bottom: "fc2_custom" | |
# bottom: "label" | |
# top: "accuracy1+2" | |
# include { | |
# phase: TEST | |
# } | |
#} | |
###Evaluation### | |
layer { | |
name: "fc1_custom" | |
type: "InnerProduct" | |
bottom: "pow1" | |
top: "fc1_custom" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
inner_product_param { | |
num_output: 100 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
name: "fc2_custom" | |
type: "InnerProduct" | |
bottom: "pow2" | |
top: "fc2_custom" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
inner_product_param { | |
num_output: 100 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
name: "fc3_custom" | |
type: "InnerProduct" | |
bottom: "pow3" | |
top: "fc3_custom" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
inner_product_param { | |
num_output: 100 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
name: "loss_1" | |
type: "SoftmaxWithLoss" | |
bottom: "fc1_custom" | |
bottom: "label" | |
top: "loss_1" | |
loss_weight: 1.0 | |
} | |
layer { | |
name: "loss_2" | |
type: "SoftmaxWithLoss" | |
bottom: "fc2_custom" | |
bottom: "label" | |
top: "loss_2" | |
loss_weight: 1.0 | |
} | |
layer { | |
name: "loss_3" | |
type: "SoftmaxWithLoss" | |
bottom: "fc3_custom" | |
bottom: "label" | |
top: "loss_3" | |
loss_weight: 1.0 | |
} | |
###ACC_Layers### | |
###Layer1### | |
layer { | |
name: "accuracy1_top-1" | |
type: "Accuracy" | |
bottom: "fc1_custom" | |
bottom: "label" | |
top: "accuracy1_top-1" | |
include { | |
phase: TEST | |
} | |
} | |
layer { | |
name: "accuracy1_top-5" | |
type: "Accuracy" | |
bottom: "fc1_custom" | |
bottom: "label" | |
top: "accuracy1_top-5" | |
accuracy_param { | |
top_k: 5 | |
} | |
include { | |
phase: TEST | |
} | |
} | |
###Layer2### | |
layer { | |
name: "accura2_top-1" | |
type: "Accuracy" | |
bottom: "fc2_custom" | |
bottom: "label" | |
top: "accuracy2_top-1" | |
include { | |
phase: TEST | |
} | |
} | |
layer { | |
name: "accuracy2_top-5" | |
type: "Accuracy" | |
bottom: "fc2_custom" | |
bottom: "label" | |
top: "accuracy2_top-5" | |
accuracy_param { | |
top_k: 5 | |
} | |
include { | |
phase: TEST | |
} | |
} | |
###Layer3### | |
layer { | |
name: "accuracy3_top-1" | |
type: "Accuracy" | |
bottom: "fc3_custom" | |
bottom: "label" | |
top: "accuracy3_top-1" | |
include { | |
phase: TEST | |
} | |
} | |
layer { | |
name: "accuracy3_top-5" | |
type: "Accuracy" | |
bottom: "fc3_custom" | |
bottom: "label" | |
top: "accuracy3_top-5" | |
accuracy_param { | |
top_k: 5 | |
} | |
include { | |
phase: TEST | |
} | |
} |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
name: "RA_CNN" | |
layer { | |
name: "data" | |
type: "Data" | |
top: "data" | |
top: "label" | |
include { | |
phase: TRAIN | |
} | |
transform_param { | |
mirror: true | |
crop_size: 448 | |
mean_value: 128 | |
mean_value: 128 | |
mean_value: 128 | |
} | |
data_param { | |
source: "/media/data/lmdb/birds" | |
batch_size: 2 | |
backend: LMDB | |
} | |
} | |
layer { | |
name: "data" | |
type: "Data" | |
top: "data" | |
top: "label" | |
include { | |
phase: TEST | |
} | |
transform_param { | |
mirror: true | |
crop_size: 448 | |
mean_value: 128 | |
mean_value: 128 | |
mean_value: 128 | |
} | |
data_param { | |
source: "/media/data/lmdb/birds" | |
batch_size: 2 | |
backend: LMDB | |
} | |
} | |
#######Scale1####### | |
layer { | |
bottom: "data" | |
top: "conv1_1" | |
name: "conv1_1" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 64 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv1_1" | |
top: "conv1_1" | |
name: "relu1_1" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv1_1" | |
top: "conv1_2" | |
name: "conv1_2" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 64 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv1_2" | |
top: "conv1_2" | |
name: "relu1_2" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv1_2" | |
top: "pool1" | |
name: "pool1" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool1" | |
top: "conv2_1" | |
name: "conv2_1" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 128 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv2_1" | |
top: "conv2_1" | |
name: "relu2_1" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv2_1" | |
top: "conv2_2" | |
name: "conv2_2" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 128 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv2_2" | |
top: "conv2_2" | |
name: "relu2_2" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv2_2" | |
top: "pool2" | |
name: "pool2" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool2" | |
top: "conv3_1" | |
name: "conv3_1" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_1" | |
top: "conv3_1" | |
name: "relu3_1" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_1" | |
top: "conv3_2" | |
name: "conv3_2" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_2" | |
top: "conv3_2" | |
name: "relu3_2" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_2" | |
top: "conv3_3" | |
name: "conv3_3" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_3" | |
top: "conv3_3" | |
name: "relu3_3" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_3" | |
top: "conv3_4" | |
name: "conv3_4" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_4" | |
top: "conv3_4" | |
name: "relu3_4" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_4" | |
top: "pool3" | |
name: "pool3" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool3" | |
top: "conv4_1" | |
name: "conv4_1" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_1" | |
top: "conv4_1" | |
name: "relu4_1" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_1" | |
top: "conv4_2" | |
name: "conv4_2" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_2" | |
top: "conv4_2" | |
name: "relu4_2" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_2" | |
top: "conv4_3" | |
name: "conv4_3" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_3" | |
top: "conv4_3" | |
name: "relu4_3" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_3" | |
top: "conv4_4" | |
name: "conv4_4" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_4" | |
top: "conv4_4" | |
name: "relu4_4" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_4" | |
top: "pool4" | |
name: "pool4" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool4" | |
top: "conv5_1" | |
name: "conv5_1" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_1" | |
top: "conv5_1" | |
name: "relu5_1" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv5_1" | |
top: "conv5_2" | |
name: "conv5_2" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_2" | |
top: "conv5_2" | |
name: "relu5_2" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv5_2" | |
top: "conv5_3" | |
name: "conv5_3" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_3" | |
top: "conv5_3" | |
name: "relu5_3" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv5_3" | |
top: "conv5_4" | |
name: "conv5_4" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_4" | |
top: "conv5_4" | |
name: "relu5_4" | |
type: "ReLU" | |
} | |
layer { | |
name: "pool5" | |
type: "Pooling" | |
bottom: "conv5_4" | |
top: "pool5" | |
pooling_param { | |
pool: AVE | |
kernel_size: 28 | |
stride: 28 | |
} | |
} | |
#######APN1####### | |
layer { | |
bottom: "conv5_4" | |
top: "anp_pool" | |
name: "anp_pool" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
name: "get_abc1" | |
type: "InnerProduct" | |
bottom: "anp_pool" | |
top: "get_abc1" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
inner_product_param { | |
num_output: 1024 | |
weight_filler { | |
type: "gaussian" | |
std: 0.001 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
name: "tanh" | |
bottom: "get_abc1" | |
top: "tanh" | |
type: "TanH" | |
} | |
layer { | |
name: "get_abc2" | |
type: "InnerProduct" | |
bottom: "tanh" | |
top: "get_abc2" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
inner_product_param { | |
num_output: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.001 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
name: "sigmoid" | |
bottom: "get_abc2" | |
top: "sig_abc" | |
type: "Sigmoid" | |
} | |
#######Scale2####### | |
layer { | |
name: "get448" | |
bottom: "sig_abc" | |
top: "get448" | |
type: "Power" | |
power_param { | |
power: 1 | |
scale: 448 | |
shift: 0 | |
} | |
} | |
layer{ | |
name: "atten_crop" | |
bottom: "data" | |
bottom: "get448" | |
top: "scale2_data" | |
type: "AttentionCrop" | |
} | |
layer { | |
bottom: "scale2_data" | |
top: "conv1_1_A" | |
name: "conv1_1_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 64 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv1_1_A" | |
top: "conv1_1_A" | |
name: "relu1_1_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv1_1_A" | |
top: "conv1_2_A" | |
name: "conv1_2_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 64 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv1_2_A" | |
top: "conv1_2_A" | |
name: "relu1_2_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv1_2_A" | |
top: "pool1_A" | |
name: "pool1_A" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool1_A" | |
top: "conv2_1_A" | |
name: "conv2_1_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 128 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv2_1_A" | |
top: "conv2_1_A" | |
name: "relu2_1_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv2_1_A" | |
top: "conv2_2_A" | |
name: "conv2_2_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 128 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv2_2_A" | |
top: "conv2_2_A" | |
name: "relu2_2_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv2_2_A" | |
top: "pool2_A" | |
name: "pool2_A" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool2_A" | |
top: "conv3_1_A" | |
name: "conv3_1_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_1_A" | |
top: "conv3_1_A" | |
name: "relu3_1_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_1_A" | |
top: "conv3_2_A" | |
name: "conv3_2_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_2_A" | |
top: "conv3_2_A" | |
name: "relu3_2_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_2_A" | |
top: "conv3_3_A" | |
name: "conv3_3_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_3_A" | |
top: "conv3_3_A" | |
name: "relu3_3_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_3_A" | |
top: "conv3_4_A" | |
name: "conv3_4_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_4_A" | |
top: "conv3_4_A" | |
name: "relu3_4_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_4_A" | |
top: "pool3_A" | |
name: "pool3_A" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool3_A" | |
top: "conv4_1_A" | |
name: "conv4_1_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_1_A" | |
top: "conv4_1_A" | |
name: "relu4_1_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_1_A" | |
top: "conv4_2_A" | |
name: "conv4_2_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_2_A" | |
top: "conv4_2_A" | |
name: "relu4_2_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_2_A" | |
top: "conv4_3_A" | |
name: "conv4_3_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_3_A" | |
top: "conv4_3_A" | |
name: "relu4_3_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_3_A" | |
top: "conv4_4_A" | |
name: "conv4_4_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_4_A" | |
top: "conv4_4_A" | |
name: "relu4_4_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_4_A" | |
top: "pool4_A" | |
name: "pool4_A" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool4_A" | |
top: "conv5_1_A" | |
name: "conv5_1_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_1_A" | |
top: "conv5_1_A" | |
name: "relu5_1_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv5_1_A" | |
top: "conv5_2_A" | |
name: "conv5_2_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_2_A" | |
top: "conv5_2_A" | |
name: "relu5_2_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv5_2_A" | |
top: "conv5_3_A" | |
name: "conv5_3_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_3_A" | |
top: "conv5_3_A" | |
name: "relu5_3_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv5_3_A" | |
top: "conv5_4_A" | |
name: "conv5_4_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_4_A" | |
top: "conv5_4_A" | |
name: "relu5_4_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv5_4_A" | |
top: "pool5_A" | |
name: "pool5_A" | |
type: "Pooling" | |
pooling_param { | |
pool: AVE | |
kernel_size: 14 | |
stride: 14 | |
} | |
} | |
#######APN2####### | |
layer { | |
name: "get_abc1_A" | |
type: "InnerProduct" | |
bottom: "conv5_4_A" | |
top: "get_abc1_A" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
inner_product_param { | |
num_output: 1024 | |
weight_filler { | |
type: "gaussian" | |
std: 0.001 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
name: "tanh_A" | |
bottom: "get_abc1_A" | |
top: "tanh_A" | |
type: "TanH" | |
} | |
layer { | |
name: "get_abc2_A" | |
type: "InnerProduct" | |
bottom: "tanh_A" | |
top: "get_abc2_A" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
inner_product_param { | |
num_output: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.001 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
name: "sigmoid_A" | |
bottom: "get_abc2_A" | |
top: "sig_abc_A" | |
type: "Sigmoid" | |
} | |
#######Scale3####### | |
layer { | |
name: "get224" | |
bottom: "sig_abc_A" | |
top: "get224" | |
type: "Power" | |
power_param { | |
power: 1 | |
scale: 224 | |
shift: 0 | |
} | |
} | |
layer{ | |
name: "atten_crop_A" | |
bottom: "scale2_data" | |
bottom: "get224" | |
top: "scale3_data" | |
type: "AttentionCrop" | |
} | |
layer { | |
bottom: "scale3_data" | |
top: "conv1_1_A_A" | |
name: "conv1_1_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 64 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv1_1_A_A" | |
top: "conv1_1_A_A" | |
name: "relu1_1_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv1_1_A_A" | |
top: "conv1_2_A_A" | |
name: "conv1_2_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 64 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv1_2_A_A" | |
top: "conv1_2_A_A" | |
name: "relu1_2_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv1_2_A_A" | |
top: "pool1_A_A" | |
name: "pool1_A_A" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool1_A_A" | |
top: "conv2_1_A_A" | |
name: "conv2_1_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 128 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv2_1_A_A" | |
top: "conv2_1_A_A" | |
name: "relu2_1_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv2_1_A_A" | |
top: "conv2_2_A_A" | |
name: "conv2_2_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 128 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv2_2_A_A" | |
top: "conv2_2_A_A" | |
name: "relu2_2_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv2_2_A_A" | |
top: "pool2_A_A" | |
name: "pool2_A_A" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool2_A_A" | |
top: "conv3_1_A_A" | |
name: "conv3_1_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_1_A_A" | |
top: "conv3_1_A_A" | |
name: "relu3_1_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_1_A_A" | |
top: "conv3_2_A_A" | |
name: "conv3_2_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_2_A_A" | |
top: "conv3_2_A_A" | |
name: "relu3_2_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_2_A_A" | |
top: "conv3_3_A_A" | |
name: "conv3_3_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_3_A_A" | |
top: "conv3_3_A_A" | |
name: "relu3_3_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_3_A_A" | |
top: "conv3_4_A_A" | |
name: "conv3_4_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_4_A_A" | |
top: "conv3_4_A_A" | |
name: "relu3_4_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_4_A_A" | |
top: "pool3_A_A" | |
name: "pool3_A_A" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool3_A_A" | |
top: "conv4_1_A_A" | |
name: "conv4_1_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_1_A_A" | |
top: "conv4_1_A_A" | |
name: "relu4_1_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_1_A_A" | |
top: "conv4_2_A_A" | |
name: "conv4_2_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_2_A_A" | |
top: "conv4_2_A_A" | |
name: "relu4_2_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_2_A_A" | |
top: "conv4_3_A_A" | |
name: "conv4_3_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_3_A_A" | |
top: "conv4_3_A_A" | |
name: "relu4_3_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_3_A_A" | |
top: "conv4_4_A_A" | |
name: "conv4_4_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_4_A_A" | |
top: "conv4_4_A_A" | |
name: "relu4_4_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_4_A_A" | |
top: "pool4_A_A" | |
name: "pool4_A_A" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool4_A_A" | |
top: "conv5_1_A_A" | |
name: "conv5_1_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_1_A_A" | |
top: "conv5_1_A_A" | |
name: "relu5_1_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv5_1_A_A" | |
top: "conv5_2_A_A" | |
name: "conv5_2_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_2_A_A" | |
top: "conv5_2_A_A" | |
name: "relu5_2_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv5_2_A_A" | |
top: "conv5_3_A_A" | |
name: "conv5_3_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_3_A_A" | |
top: "conv5_3_A_A" | |
name: "relu5_3_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv5_3_A_A" | |
top: "conv5_4_A_A" | |
name: "conv5_4_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_4_A_A" | |
top: "conv5_4_A_A" | |
name: "relu5_4_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv5_4_A_A" | |
top: "pool5_A_A" | |
name: "pool5_A_A" | |
type: "Pooling" | |
pooling_param { | |
pool: AVE | |
kernel_size: 14 | |
stride: 14 | |
} | |
} | |
#####feature_fusion##### | |
layer { | |
name: "reshape1" | |
bottom: "pool5" | |
top: "reshape1" | |
type: "Reshape" | |
reshape_param { | |
shape { | |
dim: -1 | |
dim: 512 | |
} | |
} | |
} | |
layer { | |
name: "reshape2" | |
bottom: "pool5_A" | |
top: "reshape2" | |
type: "Reshape" | |
reshape_param { | |
shape { | |
dim: -1 | |
dim: 512 | |
} | |
} | |
} | |
layer { | |
name: "reshape3" | |
bottom: "pool5_A_A" | |
top: "reshape3" | |
type: "Reshape" | |
reshape_param { | |
shape { | |
dim: -1 | |
dim: 512 | |
} | |
} | |
} | |
layer { | |
name: "pow1" | |
bottom: "reshape1" | |
top: "pow1" | |
type: "Power" | |
power_param { | |
power: 1 | |
scale: 0.1 | |
shift: 0 | |
} | |
} | |
layer { | |
name: "pow2" | |
bottom: "reshape2" | |
top: "pow2" | |
type: "Power" | |
power_param { | |
power: 1 | |
scale: 0.1 | |
shift: 0 | |
} | |
} | |
layer { | |
name: "pow3" | |
bottom: "reshape3" | |
top: "pow3" | |
type: "Power" | |
power_param { | |
power: 1 | |
scale: 0.1 | |
shift: 0 | |
} | |
} | |
#layer { | |
# name: "scale1+2+3" | |
# bottom: "pow2" | |
# bottom: "pow1" | |
# bottom: "pow3" | |
# top: "scale1+2+3" | |
# type: "Concat" | |
# concat_param { | |
# axis: 1 | |
# } | |
#} | |
#layer { | |
# name: "scale1+2" | |
# bottom: "pow2" | |
# bottom: "pow1" | |
# top: "scale1+2" | |
# type: "Concat" | |
# concat_param { | |
# axis: 1 | |
# } | |
#} | |
#layer { | |
# name: "fc1_custom" | |
# type: "InnerProduct" | |
# bottom: "scale1+2+3" | |
# top: "fc1_custom" | |
# param { | |
# lr_mult: 0.0 | |
# decay_mult: 0 | |
# } | |
# param { | |
# lr_mult: 0.0 | |
# decay_mult: 0 | |
# } | |
# inner_product_param { | |
# num_output: 100 | |
# weight_filler { | |
# type: "gaussian" | |
# std: 0.01 | |
# } | |
# bias_filler { | |
# type: "constant" | |
# value: 0 | |
# } | |
# } | |
#} | |
#layer { | |
# name: "accuracy1+2+3" | |
# type: "Accuracy" | |
# bottom: "fc1_custom" | |
# bottom: "label" | |
# top: "accuracy1+2+3" | |
# include { | |
# phase: TEST | |
# } | |
#} | |
#layer { | |
# name: "fc2_custom" | |
# type: "InnerProduct" | |
# bottom: "scale1+2" | |
# top: "fc2_custom" | |
# param { | |
# lr_mult: 0.0 | |
# decay_mult: 0 | |
# } | |
# param { | |
# lr_mult: 0.0 | |
# decay_mult: 0 | |
# } | |
# inner_product_param { | |
# num_output: 100 | |
# weight_filler { | |
# type: "gaussian" | |
# std: 0.01 | |
# } | |
# bias_filler { | |
# type: "constant" | |
# value: 0 | |
# } | |
# } | |
#} | |
#layer { | |
# name: "accuracy1+2" | |
# type: "Accuracy" | |
# bottom: "fc2_custom" | |
# bottom: "label" | |
# top: "accuracy1+2" | |
# include { | |
# phase: TEST | |
# } | |
#} | |
###Evaluation### | |
layer { | |
name: "fc1_custom" | |
type: "InnerProduct" | |
bottom: "pow1" | |
top: "fc1_custom" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
inner_product_param { | |
num_output: 100 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
name: "fc2_custom" | |
type: "InnerProduct" | |
bottom: "pow2" | |
top: "fc2_custom" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
inner_product_param { | |
num_output: 100 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
name: "fc3_custom" | |
type: "InnerProduct" | |
bottom: "pow3" | |
top: "fc3_custom" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
inner_product_param { | |
num_output: 100 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
name: "PRL_0" | |
type: "PairwiseRankingLossLayer" | |
bottom: "fc1_custom" | |
bottom: "fc2_custom" | |
top: "PRL_0" | |
loss_weight: 1.0 | |
} | |
layer { | |
name: "PRL_1" | |
type: "PairwiseRankingLossLayer" | |
bottom: "fc2_custom" | |
bottom: "fc3_custom" | |
top: "PRL_1" | |
loss_weight: 1.0 | |
} | |
###ACC_Layers### | |
###Layer1### | |
layer { | |
name: "accuracy1_top-1" | |
type: "Accuracy" | |
bottom: "fc1_custom" | |
bottom: "label" | |
top: "accuracy1_top-1" | |
include { | |
phase: TEST | |
} | |
} | |
layer { | |
name: "accuracy1_top-5" | |
type: "Accuracy" | |
bottom: "fc1_custom" | |
bottom: "label" | |
top: "accuracy1_top-5" | |
accuracy_param { | |
top_k: 5 | |
} | |
include { | |
phase: TEST | |
} | |
} | |
###Layer2### | |
layer { | |
name: "accura2_top-1" | |
type: "Accuracy" | |
bottom: "fc2_custom" | |
bottom: "label" | |
top: "accuracy2_top-1" | |
include { | |
phase: TEST | |
} | |
} | |
layer { | |
name: "accuracy2_top-5" | |
type: "Accuracy" | |
bottom: "fc2_custom" | |
bottom: "label" | |
top: "accuracy2_top-5" | |
accuracy_param { | |
top_k: 5 | |
} | |
include { | |
phase: TEST | |
} | |
} | |
###Layer3### | |
layer { | |
name: "accuracy3_top-1" | |
type: "Accuracy" | |
bottom: "fc3_custom" | |
bottom: "label" | |
top: "accuracy3_top-1" | |
include { | |
phase: TEST | |
} | |
} | |
layer { | |
name: "accuracy3_top-5" | |
type: "Accuracy" | |
bottom: "fc3_custom" | |
bottom: "label" | |
top: "accuracy3_top-5" | |
accuracy_param { | |
top_k: 5 | |
} | |
include { | |
phase: TEST | |
} | |
} |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
name: "RA_CNN" | |
#######Scale1####### | |
layer { | |
name: "data" | |
type: "Data" | |
top: "data" | |
top: "label" | |
include { | |
phase: TRAIN | |
} | |
transform_param { | |
mirror: true | |
crop_size: 448 | |
mean_value: 128 | |
mean_value: 128 | |
mean_value: 128 | |
} | |
data_param { | |
source: "/media/data/lmdb/birds" | |
batch_size: 2 | |
backend: LMDB | |
} | |
} | |
layer { | |
name: "data" | |
type: "Data" | |
top: "data" | |
top: "label" | |
include { | |
phase: TEST | |
} | |
transform_param { | |
mirror: true | |
crop_size: 448 | |
mean_value: 128 | |
mean_value: 128 | |
mean_value: 128 | |
} | |
data_param { | |
source: "/media/data/lmdb/birds" | |
batch_size: 2 | |
backend: LMDB | |
} | |
} | |
layer { | |
bottom: "data" | |
top: "conv1_1" | |
name: "conv1_1" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 64 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv1_1" | |
top: "conv1_1" | |
name: "relu1_1" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv1_1" | |
top: "conv1_2" | |
name: "conv1_2" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 64 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv1_2" | |
top: "conv1_2" | |
name: "relu1_2" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv1_2" | |
top: "pool1" | |
name: "pool1" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool1" | |
top: "conv2_1" | |
name: "conv2_1" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 128 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv2_1" | |
top: "conv2_1" | |
name: "relu2_1" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv2_1" | |
top: "conv2_2" | |
name: "conv2_2" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 128 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv2_2" | |
top: "conv2_2" | |
name: "relu2_2" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv2_2" | |
top: "pool2" | |
name: "pool2" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool2" | |
top: "conv3_1" | |
name: "conv3_1" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_1" | |
top: "conv3_1" | |
name: "relu3_1" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_1" | |
top: "conv3_2" | |
name: "conv3_2" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_2" | |
top: "conv3_2" | |
name: "relu3_2" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_2" | |
top: "conv3_3" | |
name: "conv3_3" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_3" | |
top: "conv3_3" | |
name: "relu3_3" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_3" | |
top: "conv3_4" | |
name: "conv3_4" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_4" | |
top: "conv3_4" | |
name: "relu3_4" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_4" | |
top: "pool3" | |
name: "pool3" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool3" | |
top: "conv4_1" | |
name: "conv4_1" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_1" | |
top: "conv4_1" | |
name: "relu4_1" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_1" | |
top: "conv4_2" | |
name: "conv4_2" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_2" | |
top: "conv4_2" | |
name: "relu4_2" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_2" | |
top: "conv4_3" | |
name: "conv4_3" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_3" | |
top: "conv4_3" | |
name: "relu4_3" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_3" | |
top: "conv4_4" | |
name: "conv4_4" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_4" | |
top: "conv4_4" | |
name: "relu4_4" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_4" | |
top: "pool4" | |
name: "pool4" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool4" | |
top: "conv5_1" | |
name: "conv5_1" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_1" | |
top: "conv5_1" | |
name: "relu5_1" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv5_1" | |
top: "conv5_2" | |
name: "conv5_2" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_2" | |
top: "conv5_2" | |
name: "relu5_2" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv5_2" | |
top: "conv5_3" | |
name: "conv5_3" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_3" | |
top: "conv5_3" | |
name: "relu5_3" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv5_3" | |
top: "conv5_4" | |
name: "conv5_4" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_4" | |
top: "conv5_4" | |
name: "relu5_4" | |
type: "ReLU" | |
} | |
layer { | |
name: "pool5" | |
type: "Pooling" | |
bottom: "conv5_4" | |
top: "pool5" | |
pooling_param { | |
pool: AVE | |
kernel_size: 28 | |
stride: 28 | |
} | |
} | |
#######APN1####### | |
layer { | |
bottom: "conv5_4" | |
top: "anp_pool" | |
name: "anp_pool" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
name: "get_abc1" | |
type: "InnerProduct" | |
bottom: "anp_pool" | |
top: "get_abc1" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
inner_product_param { | |
num_output: 1024 | |
weight_filler { | |
type: "gaussian" | |
std: 0.001 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
name: "tanh" | |
bottom: "get_abc1" | |
top: "tanh" | |
type: "TanH" | |
} | |
layer { | |
name: "get_abc2" | |
type: "InnerProduct" | |
bottom: "tanh" | |
top: "get_abc2" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
inner_product_param { | |
num_output: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.001 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
name: "sigmoid" | |
bottom: "get_abc2" | |
top: "sig_abc" | |
type: "Sigmoid" | |
} | |
#######Scale2####### | |
layer { | |
name: "get448" | |
bottom: "sig_abc" | |
top: "get448" | |
type: "Power" | |
power_param { | |
power: 1 | |
scale: 448 | |
shift: 0 | |
} | |
} | |
layer{ | |
name: "atten_crop" | |
bottom: "data" | |
bottom: "get448" | |
top: "scale2_data" | |
type: "AttentionCrop" | |
} | |
layer { | |
bottom: "scale2_data" | |
top: "conv1_1_A" | |
name: "conv1_1_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 64 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv1_1_A" | |
top: "conv1_1_A" | |
name: "relu1_1_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv1_1_A" | |
top: "conv1_2_A" | |
name: "conv1_2_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 64 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv1_2_A" | |
top: "conv1_2_A" | |
name: "relu1_2_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv1_2_A" | |
top: "pool1_A" | |
name: "pool1_A" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool1_A" | |
top: "conv2_1_A" | |
name: "conv2_1_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 128 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv2_1_A" | |
top: "conv2_1_A" | |
name: "relu2_1_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv2_1_A" | |
top: "conv2_2_A" | |
name: "conv2_2_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 128 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv2_2_A" | |
top: "conv2_2_A" | |
name: "relu2_2_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv2_2_A" | |
top: "pool2_A" | |
name: "pool2_A" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool2_A" | |
top: "conv3_1_A" | |
name: "conv3_1_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_1_A" | |
top: "conv3_1_A" | |
name: "relu3_1_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_1_A" | |
top: "conv3_2_A" | |
name: "conv3_2_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_2_A" | |
top: "conv3_2_A" | |
name: "relu3_2_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_2_A" | |
top: "conv3_3_A" | |
name: "conv3_3_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_3_A" | |
top: "conv3_3_A" | |
name: "relu3_3_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_3_A" | |
top: "conv3_4_A" | |
name: "conv3_4_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_4_A" | |
top: "conv3_4_A" | |
name: "relu3_4_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_4_A" | |
top: "pool3_A" | |
name: "pool3_A" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool3_A" | |
top: "conv4_1_A" | |
name: "conv4_1_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_1_A" | |
top: "conv4_1_A" | |
name: "relu4_1_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_1_A" | |
top: "conv4_2_A" | |
name: "conv4_2_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_2_A" | |
top: "conv4_2_A" | |
name: "relu4_2_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_2_A" | |
top: "conv4_3_A" | |
name: "conv4_3_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_3_A" | |
top: "conv4_3_A" | |
name: "relu4_3_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_3_A" | |
top: "conv4_4_A" | |
name: "conv4_4_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_4_A" | |
top: "conv4_4_A" | |
name: "relu4_4_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_4_A" | |
top: "pool4_A" | |
name: "pool4_A" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool4_A" | |
top: "conv5_1_A" | |
name: "conv5_1_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_1_A" | |
top: "conv5_1_A" | |
name: "relu5_1_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv5_1_A" | |
top: "conv5_2_A" | |
name: "conv5_2_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_2_A" | |
top: "conv5_2_A" | |
name: "relu5_2_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv5_2_A" | |
top: "conv5_3_A" | |
name: "conv5_3_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_3_A" | |
top: "conv5_3_A" | |
name: "relu5_3_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv5_3_A" | |
top: "conv5_4_A" | |
name: "conv5_4_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_4_A" | |
top: "conv5_4_A" | |
name: "relu5_4_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv5_4_A" | |
top: "pool5_A" | |
name: "pool5_A" | |
type: "Pooling" | |
pooling_param { | |
pool: AVE | |
kernel_size: 14 | |
stride: 14 | |
} | |
} | |
#######APN2####### | |
layer { | |
name: "get_abc1_A" | |
type: "InnerProduct" | |
bottom: "conv5_4_A" | |
top: "get_abc1_A" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
inner_product_param { | |
num_output: 1024 | |
weight_filler { | |
type: "gaussian" | |
std: 0.001 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
name: "tanh_A" | |
bottom: "get_abc1_A" | |
top: "tanh_A" | |
type: "TanH" | |
} | |
layer { | |
name: "get_abc2_A" | |
type: "InnerProduct" | |
bottom: "tanh_A" | |
top: "get_abc2_A" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
inner_product_param { | |
num_output: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.001 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
name: "sigmoid_A" | |
bottom: "get_abc2_A" | |
top: "sig_abc_A" | |
type: "Sigmoid" | |
} | |
#######Scale3####### | |
layer { | |
name: "get224" | |
bottom: "sig_abc_A" | |
top: "get224" | |
type: "Power" | |
power_param { | |
power: 1 | |
scale: 224 | |
shift: 0 | |
} | |
} | |
layer{ | |
name: "atten_crop_A" | |
bottom: "scale2_data" | |
bottom: "get224" | |
top: "scale3_data" | |
type: "AttentionCrop" | |
} | |
layer { | |
bottom: "scale3_data" | |
top: "conv1_1_A_A" | |
name: "conv1_1_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 64 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv1_1_A_A" | |
top: "conv1_1_A_A" | |
name: "relu1_1_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv1_1_A_A" | |
top: "conv1_2_A_A" | |
name: "conv1_2_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 64 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv1_2_A_A" | |
top: "conv1_2_A_A" | |
name: "relu1_2_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv1_2_A_A" | |
top: "pool1_A_A" | |
name: "pool1_A_A" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool1_A_A" | |
top: "conv2_1_A_A" | |
name: "conv2_1_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 128 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv2_1_A_A" | |
top: "conv2_1_A_A" | |
name: "relu2_1_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv2_1_A_A" | |
top: "conv2_2_A_A" | |
name: "conv2_2_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 128 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv2_2_A_A" | |
top: "conv2_2_A_A" | |
name: "relu2_2_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv2_2_A_A" | |
top: "pool2_A_A" | |
name: "pool2_A_A" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool2_A_A" | |
top: "conv3_1_A_A" | |
name: "conv3_1_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_1_A_A" | |
top: "conv3_1_A_A" | |
name: "relu3_1_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_1_A_A" | |
top: "conv3_2_A_A" | |
name: "conv3_2_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_2_A_A" | |
top: "conv3_2_A_A" | |
name: "relu3_2_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_2_A_A" | |
top: "conv3_3_A_A" | |
name: "conv3_3_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_3_A_A" | |
top: "conv3_3_A_A" | |
name: "relu3_3_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_3_A_A" | |
top: "conv3_4_A_A" | |
name: "conv3_4_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 256 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv3_4_A_A" | |
top: "conv3_4_A_A" | |
name: "relu3_4_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv3_4_A_A" | |
top: "pool3_A_A" | |
name: "pool3_A_A" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool3_A_A" | |
top: "conv4_1_A_A" | |
name: "conv4_1_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_1_A_A" | |
top: "conv4_1_A_A" | |
name: "relu4_1_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_1_A_A" | |
top: "conv4_2_A_A" | |
name: "conv4_2_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_2_A_A" | |
top: "conv4_2_A_A" | |
name: "relu4_2_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_2_A_A" | |
top: "conv4_3_A_A" | |
name: "conv4_3_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_3_A_A" | |
top: "conv4_3_A_A" | |
name: "relu4_3_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_3_A_A" | |
top: "conv4_4_A_A" | |
name: "conv4_4_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv4_4_A_A" | |
top: "conv4_4_A_A" | |
name: "relu4_4_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv4_4_A_A" | |
top: "pool4_A_A" | |
name: "pool4_A_A" | |
type: "Pooling" | |
pooling_param { | |
pool: MAX | |
kernel_size: 2 | |
stride: 2 | |
} | |
} | |
layer { | |
bottom: "pool4_A_A" | |
top: "conv5_1_A_A" | |
name: "conv5_1_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_1_A_A" | |
top: "conv5_1_A_A" | |
name: "relu5_1_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv5_1_A_A" | |
top: "conv5_2_A_A" | |
name: "conv5_2_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_2_A_A" | |
top: "conv5_2_A_A" | |
name: "relu5_2_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv5_2_A_A" | |
top: "conv5_3_A_A" | |
name: "conv5_3_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_3_A_A" | |
top: "conv5_3_A_A" | |
name: "relu5_3_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv5_3_A_A" | |
top: "conv5_4_A_A" | |
name: "conv5_4_A_A" | |
type: "Convolution" | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 0.0 | |
decay_mult: 0 | |
} | |
convolution_param { | |
num_output: 512 | |
pad: 1 | |
kernel_size: 3 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
bottom: "conv5_4_A_A" | |
top: "conv5_4_A_A" | |
name: "relu5_4_A_A" | |
type: "ReLU" | |
} | |
layer { | |
bottom: "conv5_4_A_A" | |
top: "pool5_A_A" | |
name: "pool5_A_A" | |
type: "Pooling" | |
pooling_param { | |
pool: AVE | |
kernel_size: 14 | |
stride: 14 | |
} | |
} | |
#####feature_fusion##### | |
layer { | |
name: "reshape1" | |
bottom: "pool5" | |
top: "reshape1" | |
type: "Reshape" | |
reshape_param { | |
shape { | |
dim: -1 | |
dim: 512 | |
} | |
} | |
} | |
layer { | |
name: "reshape2" | |
bottom: "pool5_A" | |
top: "reshape2" | |
type: "Reshape" | |
reshape_param { | |
shape { | |
dim: -1 | |
dim: 512 | |
} | |
} | |
} | |
layer { | |
name: "reshape3" | |
bottom: "pool5_A_A" | |
top: "reshape3" | |
type: "Reshape" | |
reshape_param { | |
shape { | |
dim: -1 | |
dim: 512 | |
} | |
} | |
} | |
layer { | |
name: "pow1" | |
bottom: "reshape1" | |
top: "pow1" | |
type: "Power" | |
power_param { | |
power: 1 | |
scale: 0.1 | |
shift: 0 | |
} | |
} | |
layer { | |
name: "pow2" | |
bottom: "reshape2" | |
top: "pow2" | |
type: "Power" | |
power_param { | |
power: 1 | |
scale: 0.1 | |
shift: 0 | |
} | |
} | |
layer { | |
name: "pow3" | |
bottom: "reshape3" | |
top: "pow3" | |
type: "Power" | |
power_param { | |
power: 1 | |
scale: 0.1 | |
shift: 0 | |
} | |
} | |
layer { | |
name: "scale1+2+3" | |
bottom: "pow2" | |
bottom: "pow1" | |
bottom: "pow3" | |
top: "scale1+2+3" | |
type: "Concat" | |
concat_param { | |
axis: 1 | |
} | |
} | |
layer { | |
name: "scale1+2" | |
bottom: "pow2" | |
bottom: "pow1" | |
top: "scale1+2" | |
type: "Concat" | |
concat_param { | |
axis: 1 | |
} | |
} | |
layer { | |
name: "fc1_custom" | |
type: "InnerProduct" | |
bottom: "scale1+2+3" | |
top: "fc1_custom" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
inner_product_param { | |
num_output: 100 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
name: "accuracy1+2+3" | |
type: "Accuracy" | |
bottom: "fc1_custom" | |
bottom: "label" | |
top: "accuracy1+2+3" | |
include { | |
phase: TEST | |
} | |
} | |
layer { | |
name: "fc2_custom" | |
type: "InnerProduct" | |
bottom: "scale1+2" | |
top: "fc2_custom" | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
param { | |
lr_mult: 1.0 | |
decay_mult: 0 | |
} | |
inner_product_param { | |
num_output: 100 | |
weight_filler { | |
type: "gaussian" | |
std: 0.01 | |
} | |
bias_filler { | |
type: "constant" | |
value: 0 | |
} | |
} | |
} | |
layer { | |
name: "accuracy1+2" | |
type: "Accuracy" | |
bottom: "fc2_custom" | |
bottom: "label" | |
top: "accuracy1+2" | |
include { | |
phase: TEST | |
} | |
} | |
layer { | |
name: "loss_1" | |
type: "SoftmaxWithLoss" | |
bottom: "fc2_custom" | |
bottom: "label" | |
top: "loss_1" | |
loss_weight: 1.0 | |
} | |
layer { | |
name: "loss_0" | |
type: "SoftmaxWithLoss" | |
bottom: "fc1_custom" | |
bottom: "label" | |
top: "loss_0" | |
loss_weight: 1.0 | |
} |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment