我正在处理使用data来生成图像输出的情况，我知道我可以通过定义两个数据层来实现这一点，其中一个顶部是{{1} }和其他输出是label。我使用EUCLIDEAN_LOSS作为我的损失函数。 data和label都是从leveldb转换的图像数据。我构建的网络如下所示：
name: "Net"
layer {
name: "data"
type: "DenseImageData"
top: "data"
top: "label"
dense_image_data_param {
source: "/home/path/trainDataSet.txt"
batch_size: 1
shuffle: true
}
}
# ***** conv1 *****
layer {
name: "conv1_1"
type: "Convolution"
bottom: "data"
top: "conv1_1"
convolution_param {
num_output: 64
pad: 1
kernel_size: 3
}
}
layer {
name: "relu1_1"
type: "ReLU"
bottom: "conv1_1"
top: "conv1_1"
}
layer {
name: "conv1_2"
type: "Convolution"
bottom: "conv1_1"
top: "conv1_2"
convolution_param {
num_output: 64
pad: 1
kernel_size: 3
stride: 2
}
}
layer {
name: "relu1_2"
type: "ReLU"
bottom: "conv1_2"
top: "conv1_2"
}
layer {
bottom: "conv1_2"
top: "pool1"
top: "pool1_mask"
name: "pool1"
type: "Pooling"
pooling_param {
pool: MAX
kernel_size: 2
stride: 2
}
}
# ***** conv2 *****
layer {
name: "conv2_1"
type: "Convolution"
bottom: "pool1"
top: "conv2_1"
convolution_param {
num_output: 128
pad: 1
kernel_size: 3
}
}
layer {
name: "relu2_1"
type: "ReLU"
bottom: "conv2_1"
top: "conv2_1"
}
layer {
name: "conv2_2"
type: "Convolution"
bottom: "conv2_1"
top: "conv2_2"
convolution_param {
num_output: 128
pad: 1
kernel_size: 3
stride: 2
}
}
layer {
name: "relu2_2"
type: "ReLU"
bottom: "conv2_2"
top: "conv2_2"
}
layer {
bottom: "conv2_2"
top: "pool2"
top: "pool2_mask"
name: "pool2"
type: "Pooling"
pooling_param {
pool: MAX
kernel_size: 2
stride: 2
}
}
# ***** conv3 *****
layer {
name: "conv3_1"
type: "Convolution"
bottom: "pool2"
top: "conv3_1"
convolution_param {
num_output: 256
pad: 1
kernel_size: 3
}
}
layer {
name: "relu3_1"
type: "ReLU"
bottom: "conv3_1"
top: "conv3_1"
}
layer {
name: "conv3_2"
type: "Convolution"
bottom: "conv3_1"
top: "conv3_2"
convolution_param {
num_output: 256
pad: 1
kernel_size: 3
}
}
layer {
name: "relu3_2"
type: "ReLU"
bottom: "conv3_2"
top: "conv3_2"
}
layer {
name: "conv3_3"
type: "Convolution"
bottom: "conv3_2"
top: "conv3_3"
convolution_param {
num_output: 256
pad: 1
kernel_size: 3
stride: 2
}
}
layer {
name: "relu3_3"
type: "ReLU"
bottom: "conv3_3"
top: "conv3_3"
}
layer {
bottom: "conv3_3"
top: "pool3"
top: "pool3_mask"
name: "pool3"
type: "Pooling"
pooling_param {
pool: MAX
kernel_size: 2
stride: 2
}
}
# ***** conv4 *****
layer {
name: "conv4_1"
type: "Convolution"
bottom: "pool3"
top: "conv4_1"
convolution_param {
num_output: 512
kernel_size: 3
stride: 1
pad: 1
# dilation: 1
}
}
layer {
name: "relu4_1"
type: "ReLU"
bottom: "conv4_1"
top: "conv4_1"
}
layer {
name: "conv4_2"
type: "Convolution"
bottom: "conv4_1"
top: "conv4_2"
convolution_param {
num_output: 512
kernel_size: 3
stride: 1
pad: 1
# dilation: 1
}
}
layer {
name: "relu4_2"
type: "ReLU"
bottom: "conv4_2"
top: "conv4_2"
}
layer {
name: "conv4_3"
type: "Convolution"
bottom: "conv4_2"
top: "conv4_3"
convolution_param {
num_output: 512
kernel_size: 3
stride: 1
pad: 1
# dilation: 1
}
}
layer {
name: "relu4_3"
type: "ReLU"
bottom: "conv4_3"
top: "conv4_3"
}
layer {
bottom: "conv4_3"
top: "pool4"
top: "pool4_mask"
name: "pool4"
type: "Pooling"
pooling_param {
pool: MAX
kernel_size: 2
stride: 2
}
}
# ***** conv5 *****
layer {
name: "conv5_1"
type: "Convolution"
bottom: "pool4"
top: "conv5_1"
convolution_param {
num_output: 512
kernel_size: 3
stride: 1
pad: 2
# dilation: 2
}
}
layer {
name: "relu5_1"
type: "ReLU"
bottom: "conv5_1"
top: "conv5_1"
}
layer {
name: "conv5_2"
type: "Convolution"
bottom: "conv5_1"
top: "conv5_2"
convolution_param {
num_output: 512
kernel_size: 3
stride: 1
pad: 2
# dilation: 2
}
}
layer {
name: "relu5_2"
type: "ReLU"
bottom: "conv5_2"
top: "conv5_2"
}
layer {
name: "conv5_3"
type: "Convolution"
bottom: "conv5_2"
top: "conv5_3"
convolution_param {
num_output: 512
kernel_size: 3
stride: 1
pad: 2
# dilation: 2
}
}
layer {
name: "relu5_3"
type: "ReLU"
bottom: "conv5_3"
top: "conv5_3"
}
layer {
bottom: "conv5_3"
top: "pool5"
top: "pool5_mask"
name: "pool5"
type: "Pooling"
pooling_param {
pool: MAX
kernel_size: 2
stride: 2
}
}
# ***** Upsample 5 Layer *****
layer {
name: "upsample5"
type: "Upsample"
bottom: "pool5"
top: "pool5_D"
bottom: "pool5_mask"
upsample_param {
scale: 2
upsample_w: 2
upsample_h: 2
}
}
layer {
bottom: "pool5_D"
top: "conv5_3_D"
name: "conv5_3_D"
type: "Convolution"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 2
decay_mult: 0
}
convolution_param {
weight_filler {
type: "msra"
}
bias_filler {
type: "constant"
}
num_output: 512
pad: 1
kernel_size: 3
}
}
layer {
bottom: "conv5_3_D"
top: "conv5_3_D"
name: "conv5_3_D_bn"
type: "BN"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 1
decay_mult: 0
}
bn_param {
scale_filler {
type: "constant"
value: 1
}
shift_filler {
type: "constant"
value: 0.001
}
}
}
layer {
bottom: "conv5_3_D"
top: "conv5_3_D"
name: "relu5_3_D"
type: "ReLU"
}
layer {
bottom: "conv5_3_D"
top: "conv5_2_D"
name: "conv5_2_D"
type: "Convolution"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 2
decay_mult: 0
}
convolution_param {
weight_filler {
type: "msra"
}
bias_filler {
type: "constant"
}
num_output: 512
pad: 1
kernel_size: 3
}
}
layer {
bottom: "conv5_2_D"
top: "conv5_2_D"
name: "conv5_2_D_bn"
type: "BN"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 1
decay_mult: 0
}
bn_param {
scale_filler {
type: "constant"
value: 1
}
shift_filler {
type: "constant"
value: 0.001
}
}
}
layer {
bottom: "conv5_2_D"
top: "conv5_2_D"
name: "relu5_2_D"
type: "ReLU"
}
layer {
bottom: "conv5_2_D"
top: "conv5_1_D"
name: "conv5_1_D"
type: "Convolution"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 2
decay_mult: 0
}
convolution_param {
weight_filler {
type: "msra"
}
bias_filler {
type: "constant"
}
num_output: 512
pad: 1
kernel_size: 3
}
}
layer {
bottom: "conv5_1_D"
top: "conv5_1_D"
name: "conv5_1_D_bn"
type: "BN"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 1
decay_mult: 0
}
bn_param {
scale_filler {
type: "constant"
value: 1
}
shift_filler {
type: "constant"
value: 0.001
}
}
}
layer {
bottom: "conv5_1_D"
top: "conv5_1_D"
name: "relu5_1_D"
type: "ReLU"
}
# ***** Upsample 4 Layer *****
layer {
name: "upsample4"
type: "Upsample"
bottom: "conv5_1_D"
top: "pool4_D"
bottom: "pool4_mask"
upsample_param {
scale: 2
upsample_w: 3
upsample_h: 3
}
}
layer {
bottom: "pool4_D"
top: "conv4_3_D"
name: "conv4_3_D"
type: "Convolution"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 2
decay_mult: 0
}
convolution_param {
weight_filler {
type: "msra"
}
bias_filler {
type: "constant"
}
num_output: 512
pad: 1
kernel_size: 3
}
}
layer {
bottom: "conv4_3_D"
top: "conv4_3_D"
name: "conv4_3_D_bn"
type: "BN"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 1
decay_mult: 0
}
bn_param {
scale_filler {
type: "constant"
value: 1
}
shift_filler {
type: "constant"
value: 0.001
}
}
}
layer {
bottom: "conv4_3_D"
top: "conv4_3_D"
name: "relu4_3_D"
type: "ReLU"
}
layer {
bottom: "conv4_3_D"
top: "conv4_2_D"
name: "conv4_2_D"
type: "Convolution"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 2
decay_mult: 0
}
convolution_param {
weight_filler {
type: "msra"
}
bias_filler {
type: "constant"
}
num_output: 512
pad: 1
kernel_size: 3
}
}
layer {
bottom: "conv4_2_D"
top: "conv4_2_D"
name: "conv4_2_D_bn"
type: "BN"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 1
decay_mult: 0
}
bn_param {
scale_filler {
type: "constant"
value: 1
}
shift_filler {
type: "constant"
value: 0.001
}
}
}
layer {
bottom: "conv4_2_D"
top: "conv4_2_D"
name: "relu4_2_D"
type: "ReLU"
}
layer {
bottom: "conv4_2_D"
top: "conv4_1_D"
name: "conv4_1_D"
type: "Convolution"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 2
decay_mult: 0
}
convolution_param {
weight_filler {
type: "msra"
}
bias_filler {
type: "constant"
}
num_output: 256
pad: 1
kernel_size: 3
}
}
layer {
bottom: "conv4_1_D"
top: "conv4_1_D"
name: "conv4_1_D_bn"
type: "BN"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 1
decay_mult: 0
}
bn_param {
scale_filler {
type: "constant"
value: 1
}
shift_filler {
type: "constant"
value: 0.001
}
}
}
layer {
bottom: "conv4_1_D"
top: "conv4_1_D"
name: "relu4_1_D"
type: "ReLU"
}
# ***** Upsample 3 Layer *****
layer {
name: "upsample3"
type: "Upsample"
bottom: "conv4_1_D"
top: "pool3_D"
bottom: "pool3_mask"
upsample_param {
upsample_w: 11
upsample_h: 11
scale: 2
}
}
layer {
bottom: "pool3_D"
top: "conv3_3_D"
name: "conv3_3_D"
type: "Convolution"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 2
decay_mult: 0
}
convolution_param {
weight_filler {
type: "msra"
}
bias_filler {
type: "constant"
}
num_output: 256
pad: 1
kernel_size: 3
}
}
layer {
bottom: "conv3_3_D"
top: "conv3_3_D"
name: "conv3_3_D_bn"
type: "BN"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 1
decay_mult: 0
}
bn_param {
scale_filler {
type: "constant"
value: 1
}
shift_filler {
type: "constant"
value: 0.001
}
}
}
layer {
bottom: "conv3_3_D"
top: "conv3_3_D"
name: "relu3_3_D"
type: "ReLU"
}
layer {
bottom: "conv3_3_D"
top: "conv3_2_D"
name: "conv3_2_D"
type: "Convolution"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 2
decay_mult: 0
}
convolution_param {
weight_filler {
type: "msra"
}
bias_filler {
type: "constant"
}
num_output: 256
pad: 1
kernel_size: 3
}
}
layer {
bottom: "conv3_2_D"
top: "conv3_2_D"
name: "conv3_2_D_bn"
type: "BN"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 1
decay_mult: 0
}
bn_param {
scale_filler {
type: "constant"
value: 1
}
shift_filler {
type: "constant"
value: 0.001
}
}
}
layer {
bottom: "conv3_2_D"
top: "conv3_2_D"
name: "relu3_2_D"
type: "ReLU"
}
layer {
bottom: "conv3_2_D"
top: "conv3_1_D"
name: "conv3_1_D"
type: "Convolution"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 2
decay_mult: 0
}
convolution_param {
weight_filler {
type: "msra"
}
bias_filler {
type: "constant"
}
num_output: 128
pad: 1
kernel_size: 3
}
}
layer {
bottom: "conv3_1_D"
top: "conv3_1_D"
name: "conv3_1_D_bn"
type: "BN"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 1
decay_mult: 0
}
bn_param {
scale_filler {
type: "constant"
value: 1
}
shift_filler {
type: "constant"
value: 0.001
}
}
}
layer {
bottom: "conv3_1_D"
top: "conv3_1_D"
name: "relu3_1_D"
type: "ReLU"
}
# ***** Upsample 2 Layer *****
layer {
name: "upsample2"
type: "Upsample"
bottom: "conv3_1_D"
top: "pool2_D"
bottom: "pool2_mask"
upsample_param {
upsample_w: 22
upsample_h: 22
scale: 2
}
}
layer {
bottom: "pool2_D"
top: "conv2_2_D"
name: "conv2_2_D"
type: "Convolution"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 2
decay_mult: 0
}
convolution_param {
weight_filler {
type: "msra"
}
bias_filler {
type: "constant"
}
num_output: 128
pad: 1
kernel_size: 3
}
}
layer {
bottom: "conv2_2_D"
top: "conv2_2_D"
name: "conv2_2_D_bn"
type: "BN"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 1
decay_mult: 0
}
bn_param {
scale_filler {
type: "constant"
value: 1
}
shift_filler {
type: "constant"
value: 0.001
}
}
}
layer {
bottom: "conv2_2_D"
top: "conv2_2_D"
name: "relu2_2_D"
type: "ReLU"
}
layer {
bottom: "conv2_2_D"
top: "conv2_1_D"
name: "conv2_1_D"
type: "Convolution"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 2
decay_mult: 0
}
convolution_param {
weight_filler {
type: "msra"
}
bias_filler {
type: "constant"
}
num_output: 64
pad: 1
kernel_size: 3
}
}
layer {
bottom: "conv2_1_D"
top: "conv2_1_D"
name: "conv2_1_D_bn"
type: "BN"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 1
decay_mult: 0
}
bn_param {
scale_filler {
type: "constant"
value: 1
}
shift_filler {
type: "constant"
value: 0.001
}
}
}
layer {
bottom: "conv2_1_D"
top: "conv2_1_D"
name: "relu2_1_D"
type: "ReLU"
}
# ***** Upsample 1 Layer *****
layer {
name: "upsample1"
type: "Upsample"
bottom: "conv2_1_D"
top: "pool1_D"
bottom: "pool1_mask"
upsample_param {
upsample_w: 44
upsample_h: 44
scale: 2
}
}
layer {
bottom: "pool1_D"
top: "conv1_2_D"
name: "conv1_2_D"
type: "Convolution"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 2
decay_mult: 0
}
convolution_param {
weight_filler {
type: "msra"
}
bias_filler {
type: "constant"
}
num_output: 64
pad: 1
kernel_size: 3
}
}
layer {
bottom: "conv1_2_D"
top: "conv1_2_D"
name: "conv1_2_D_bn"
type: "BN"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 1
decay_mult: 0
}
bn_param {
scale_filler {
type: "constant"
value: 1
}
shift_filler {
type: "constant"
value: 0.001
}
}
}
layer {
bottom: "conv1_2_D"
top: "conv1_2_D"
name: "relu1_2_D"
type: "ReLU"
}
layer {
bottom: "conv1_2_D"
top: "conv1_1_D"
name: "conv1_1_D"
type: "Convolution"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 2
decay_mult: 0
}
convolution_param {
weight_filler {
type: "msra"
}
bias_filler {
type: "constant"
}
num_output: 30976
pad: 1
kernel_size: 3
}
}
## ***** Output Layer *****
#
#layer {
# name: "conv5_176"
# type: "Convolution"
# bottom: "conv5_3"
# top: "conv8_176"
# convolution_param {
# num_output: 30976 #176X176
# kernel_size: 1
# stride: 1
# # dilation: 1
# }
#}
# ***** Softmax loss *****
layer {
name: "loss"
type: "SoftmaxWithLoss"
bottom: "conv1_1_D"
bottom: "label"
top: "loss"
softmax_param {engine: CAFFE}
loss_weight: 1.0
}
问题是，当我训练网络时，我对问题的解释如下：
I1127 14:13:15.155333 36915 net.cpp:90] Creating Layer upsample1
I1127 14:13:15.155349 36915 net.cpp:410] upsample1 <- conv2_1_D
I1127 14:13:15.155360 36915 net.cpp:410] upsample1 <- pool1_mask
I1127 14:13:15.155375 36915 net.cpp:368] upsample1 -> pool1_D
I1127 14:13:15.155395 36915 net.cpp:120] Setting up upsample1
F1127 14:13:15.155416 36915 upsample_layer.cpp:63] Check failed: bottom[0]->height() == bottom[1]->height() (22 vs. 44)
它表明损失层的两个底部是相同的通道。
有人可以就如何解决问题提出一些建议吗？