Skip to first unread message
p.Paul
Feb 15, 2017, 1:03:24 PM2/15/17
to Caffe Users
Trying to do a regression task, having 48 outputs per image.
The following is the deploy.prototxt
Is there any one, has done similar problem?
I get some values till fc7, but at lp_fc8, it gives 0 values.
name: "grl"
# ---------------------------------------------------------------------- Source
# Train phase
# synthetic images
layer {
name: "image_data"
type: "ImageData"
top: "data"
top: "label1"
include {
phase: TRAIN
}
transform_param {
scale: 0.00390625
}
image_data_param {
source: "data/imagenet/train/images.txt"
root_folder: "data/imagenet/train/"
batch_size: 50
is_color: false
shuffle: false
}
}
# pose, or other labels
layer {
name: "source_label"
type: "Data"
top: "lp_labels"
data_param {
source: "data/label-lmdb_test"
backend: LMDB
batch_size: 50
}
include: { phase: TRAIN }
}
# ---------------------------------------------------------------------- Test
# Test phase
layer {
name: "image_data"
type: "ImageData"
top: "data"
top: "label1"
include {
phase: TEST
}
transform_param {
scale: 0.00390625
}
image_data_param {
source: "data/imagenet/train/images.txt"
root_folder: "data/imagenet/train/"
batch_size: 1
is_color: false
shuffle: false
}
}
layer {
name: "valid_label"
type: "Data"
top: "lp_labels"
data_param {
source: "data/label-lmdb_test"
backend: LMDB
batch_size: 1
}
include: { phase: TEST }
}
# ----------------------------------------------------------------------------
# is_grl_layer 2 means that it is feature layer
layer {
name: "conv1"
type: "Convolution"
bottom: "data"
top: "conv1"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 2
decay_mult: 0
}
convolution_param {
num_output: 96
kernel_size: 11
stride: 4
weight_filler {
type: "gaussian"
std: 0.01
}
bias_filler {
type: "constant"
value: 0
}
}
}
layer {
name: "relu1"
type: "ReLU"
bottom: "conv1"
top: "conv1"
}
layer {
name: "norm1"
type: "LRN"
bottom: "conv1"
top: "norm1"
lrn_param {
local_size: 5
alpha: 0.0001
beta: 0.75
}
}
layer {
name: "pool1"
type: "Pooling"
bottom: "norm1"
top: "pool1"
pooling_param {
pool: MAX
kernel_size: 3
stride: 2
}
}
# ----------------------------------------------------------------------------
layer {
name: "conv2"
type: "Convolution"
bottom: "pool1"
top: "conv2"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 2
decay_mult: 0
}
convolution_param {
num_output: 256
pad: 2
kernel_size: 5
group: 2
weight_filler {
type: "gaussian"
std: 0.01
}
bias_filler {
type: "constant"
value: 0.1
}
}
}
layer {
name: "relu2"
type: "ReLU"
bottom: "conv2"
top: "conv2"
}
layer {
name: "norm2"
type: "LRN"
bottom: "conv2"
top: "norm2"
lrn_param {
local_size: 5
alpha: 0.0001
beta: 0.75
}
}
layer {
name: "pool2"
type: "Pooling"
bottom: "norm2"
top: "pool2"
pooling_param {
pool: MAX
kernel_size: 3
stride: 2
}
}
# ----------------------------------------------------------------------------
layer {
name: "conv3"
type: "Convolution"
bottom: "pool2"
top: "conv3"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 2
decay_mult: 0
}
convolution_param {
num_output: 384
pad: 1
kernel_size: 3
weight_filler {
type: "gaussian"
std: 0.01
}
bias_filler {
type: "constant"
value: 0
}
}
}
layer {
name: "relu3"
type: "ReLU"
bottom: "conv3"
top: "conv3"
}
# ----------------------------------------------------------------------------
layer {
name: "conv4"
type: "Convolution"
bottom: "conv3"
top: "conv4"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 2
decay_mult: 0
}
convolution_param {
num_output: 384
pad: 1
kernel_size: 3
group: 2
weight_filler {
type: "gaussian"
std: 0.01
}
bias_filler {
type: "constant"
value: 0.1
}
}
}
layer {
name: "relu4"
type: "ReLU"
bottom: "conv4"
top: "conv4"
}
# ----------------------------------------------------------------------------
layer {
name: "conv5"
type: "Convolution"
bottom: "conv4"
top: "conv5"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 2
decay_mult: 0
}
convolution_param {
num_output: 256
pad: 1
kernel_size: 3
group: 2
weight_filler {
type: "gaussian"
std: 0.01
}
bias_filler {
type: "constant"
value: 0.1
}
}
}
layer {
name: "relu5"
type: "ReLU"
bottom: "conv5"
top: "conv5"
}
layer {
name: "pool5"
type: "Pooling"
bottom: "conv5"
top: "pool5"
pooling_param {
pool: MAX
kernel_size: 3
stride: 2
}
}
layer {
name: "fc6"
type: "InnerProduct"
bottom: "pool5"
top: "fc6"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 2
decay_mult: 0
}
inner_product_param {
num_output: 345
weight_filler {
type: "gaussian"
std: 0.005
}
bias_filler {
type: "constant"
value: 0.1
}
}
}
layer {
name: "relu6"
type: "ReLU"
bottom: "fc6"
top: "fc6"
}
layer {
name: "drop6"
type: "Dropout"
bottom: "fc6"
top: "fc6"
dropout_param {
dropout_ratio: 0.5
}
}
# ----------------------------------------------------------------------------
layer {
name: "fc7"
type: "InnerProduct"
bottom: "fc6"
top: "fc7"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 2
decay_mult: 0
}
inner_product_param {
num_output: 4096
weight_filler {
type: "gaussian"
std: 0.005
}
bias_filler {
type: "constant"
value: 0.1
}
}
}
layer {
name: "relu7"
type: "ReLU"
bottom: "fc7"
top: "fc7"
}
layer {
name: "drop7"
type: "Dropout"
bottom: "fc7"
top: "fc7"
dropout_param {
dropout_ratio: 0.5
}
}
# ----------------------------------------------------------------------------
# -----------------------------------------------------------------------------
# ------------------------------------------------------------- Label predictor
# -----------------------------------------------------------------------------
# ----------------------------------------------------------------------------
layer {
name: "lp_fc8"
type: "InnerProduct"
bottom: "fc7"
top: "lp_fc8"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 2
decay_mult: 0
}
inner_product_param {
num_output:48
weight_filler {
type: "gaussian"
std: 0.01
}
bias_filler {
type: "constant"
value: 0
}
}
}
#layer {
#name: "sigmoid8"
#type: "Sigmoid"
#bottom: "lp_fc8"
#top: "predict"
#}
# is_grl_loss 0 means that it is regressor loss.
# it will back propagation anyway.
layer {
name: "lp_loss"
type: "EuclideanLoss"
bottom: "lp_fc8"
bottom: "lp_labels"
top: "lp_loss"
loss_weight: 2
}
The following is the deploy.prototxt
name: "grl"
input: "data"
input_dim: 1
input_dim: 1
input_dim: 270
input_dim: 480
layer {
name: "conv1"
type: "Convolution"
bottom: "data"
top: "conv1"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 2
decay_mult: 0
}
convolution_param {
num_output: 96
kernel_size: 11
stride: 4
}
}
layer {
name: "relu1"
type: "ReLU"
bottom: "conv1"
top: "conv1"
}
layer {
name: "pool1"
type: "Pooling"
bottom: "conv1"
top: "pool1"
pooling_param {
pool: MAX
kernel_size: 3
stride: 2
}
}
layer {
name: "norm1"
type: "LRN"
bottom: "pool1"
top: "norm1"
lrn_param {
local_size: 5
alpha: 0.0001
beta: 0.75
}
}
layer {
name: "conv2"
type: "Convolution"
bottom: "norm1"
top: "conv2"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 2
decay_mult: 0
}
convolution_param {
num_output: 256
pad: 2
kernel_size: 5
group: 2
}
}
layer {
name: "relu2"
type: "ReLU"
bottom: "conv2"
top: "conv2"
}
layer {
name: "pool2"
type: "Pooling"
bottom: "conv2"
top: "pool2"
pooling_param {
pool: MAX
kernel_size: 3
stride: 2
}
}
layer {
name: "norm2"
type: "LRN"
bottom: "pool2"
top: "norm2"
lrn_param {
local_size: 5
alpha: 0.0001
beta: 0.75
}
}
layer {
name: "conv3"
type: "Convolution"
bottom: "norm2"
top: "conv3"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 2
decay_mult: 0
}
convolution_param {
num_output: 384
pad: 1
kernel_size: 3
}
}
layer {
name: "relu3"
type: "ReLU"
bottom: "conv3"
top: "conv3"
}
layer {
name: "conv4"
type: "Convolution"
bottom: "conv3"
top: "conv4"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 2
decay_mult: 0
}
convolution_param {
num_output: 384
pad: 1
kernel_size: 3
group: 2
}
}
layer {
name: "relu4"
type: "ReLU"
bottom: "conv4"
top: "conv4"
}
layer {
name: "conv5"
type: "Convolution"
bottom: "conv4"
top: "conv5"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 2
decay_mult: 0
}
convolution_param {
num_output: 256
pad: 1
kernel_size: 3
group: 2
}
}
layer {
name: "relu5"
type: "ReLU"
bottom: "conv5"
top: "conv5"
}
layer {
name: "pool5"
type: "Pooling"
bottom: "conv5"
top: "pool5"
pooling_param {
pool: MAX
kernel_size: 3
stride: 2
}
}
layer {
name: "fc6"
type: "InnerProduct"
bottom: "pool5"
top: "fc6"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 2
decay_mult: 0
}
inner_product_param {
num_output: 345
}
}
layer {
name: "relu6"
type: "ReLU"
bottom: "fc6"
top: "fc6"
}
layer {
name: "drop6"
type: "Dropout"
bottom: "fc6"
top: "fc6"
dropout_param {
dropout_ratio: 0.5
}
}
layer {
name: "fc7"
type: "InnerProduct"
bottom: "fc6"
top: "fc7"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 2
decay_mult: 0
}
inner_product_param {
num_output: 4096
}
}
layer {
name: "relu7"
type: "ReLU"
bottom: "fc7"
top: "fc7"
}
layer {
name: "drop7"
type: "Dropout"
bottom: "fc7"
top: "fc7"
dropout_param {
dropout_ratio: 0.5
}
}
layer {
name: "my_fc8"
type: "InnerProduct"
bottom: "fc7"
top: "my_fc8"
param {
lr_mult: 1
decay_mult: 1
}
param {
lr_mult: 20
decay_mult: 0
}
inner_product_param {
num_output: 48
}
}
#layer {
#name: "sigmoid8"
#type: "Sigmoid"
#bottom: "my_fc8"
#top: "predict"
#}
Enter code here...Is there any one, has done similar problem?
kishen suraj P
Feb 15, 2017, 9:56:56 PM2/15/17
to Caffe Users
Zero values means shape of the output is zero?
p.Paul
Feb 16, 2017, 3:50:45 AM2/16/17
to Caffe Users
No,
Both these matlab lines gives me 0 values , params and data
fc8 = net.blobs('my_fc8').get_data()
fc8_param = net.params('my_fc8',1).get_data();
Both these matlab lines gives me 0 values , params and data
fc8 = net.blobs('my_fc8').get_data()
fc8_param = net.params('my_fc8',1).get_data();
p.Paul
Feb 16, 2017, 3:54:12 AM2/16/17
to Caffe Users
I get some values till fc7, but at my_fc8, it gives 0 values. Imean Output values , shape is not zero. It is 50(batch size) x 1 x 1x 48
On Thursday, February 16, 2017 at 3:56:56 AM UTC+1, kishen suraj P wrote:
p.Paul
Feb 16, 2017, 11:59:23 AM2/16/17
to Caffe Users
I have no clue about this problem. Can someone give at least some thoughts on it.I am really stuck
par...@unlv.nevada.edu
Feb 16, 2017, 3:59:50 PM2/16/17
to Caffe Users
Most probably your network is way bigger than what you need (I am guessing your dataset is very small compared to your network size).
Try a smaller network - one with less layers and/or less outputs for each layer.
Try a smaller network - one with less layers and/or less outputs for each layer.
Reply all
Reply to author
Forward
0 new messages