LSTM clip input

263 views

Skip to first unread message

Yasin Almalıoğlu

unread,

Aug 9, 2017, 6:47:35 AM8/9/17

to Caffe Users

I am trying to construct a model with LSTM layer for image regression on top of several convolutional layers. LSTM layer in the Caffe release requires 2 input blobs. I have problems with the clip input (c0) of LSTM layer. I have two consecutive LSTM layers, lstm1 and lstm2. I manually adjust the shape of the first clip input to the lstm1 and lstm1 accepts it without any problem. However, the second lstm2 does not accept it giving an error:

F0809 13:11:56.901882  3608 pooling_layer.cpp:81] Check failed: 4 == bottom[0]->num_axes() (4 vs. 3) Input must have 4 axes, corresponding to (num, channels, height, width)

How can I give clip inputs to LSTM layers? There are many PRs on the web bu non is the same as the one in the Caffe release.

this is my model:

input: "clip"
input_shape { dim: 1 dim: 51 }

layer {
  name: "data"
  type: "Data"
  top: "data"
  data_param {
    source: "/home/yasin/lmdb_dataset/train_lmdb"
    batch_size: 1
    backend: LMDB
  }
}
layer {
  name: "label"
  type: "Data"
  top: "label"
  data_param {
    source: "/home/yasin/lmdb_dataset/train_label_lmdb"
    batch_size: 1
    backend: LMDB
  }
}


layer {
  name: "label_wpqr"
  type: "Slice"
  bottom: "label"
  top: "label_wpqr"
  top: "label_xyz"
  slice_param {
    slice_point: 4
    axis: 1
  }
}
layer {
  name: "conv1"
  type: "Convolution"
  bottom: "data"
  top: "conv1"
  convolution_param {
    num_output: 64
    pad: 3
    kernel_size: 7
    stride: 2
    weight_filler {
      type: "gaussian"
      std: 0.0149999996647
    }
    bias_filler {
      type: "constant"
      value: 0.0
    }
  }
}
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
    pad: 0
  }
}
layer {
  name: "norm1"
  type: "LRN"
  bottom: "pool1"
  top: "norm1"
  lrn_param {
    local_size: 5
    alpha: 9.99999974738e-05
    beta: 0.75
  }
}
layer {
  name: "reduction2"
  type: "Convolution"
  bottom: "norm1"
  top: "reduction2"
  convolution_param {
    num_output: 64
    pad: 0
    kernel_size: 1
    group: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0.0
    }
  }
}
layer {
  name: "relu_reduction2"
  type: "ReLU"
  bottom: "reduction2"
  top: "reduction2"
}
layer {
  name: "conv2"
  type: "Convolution"
  bottom: "conv1"
  top: "conv2"
  convolution_param {
    num_output: 192
    pad: 1
    kernel_size: 3
    group: 1
    weight_filler {
      type: "gaussian"
      std: 0.019999999553
    }
    bias_filler {
      type: "constant"
      value: 0.0
    }
  }
}
layer {
  name: "relu2"
  type: "ReLU"
  bottom: "conv2"
  top: "conv2"
}
layer {
  name: "norm2"
  type: "LRN"
  bottom: "conv2"
  top: "norm2"
  lrn_param {
    local_size: 5
    alpha: 9.99999974738e-05
    beta: 0.75
  }
}
layer {
  name: "pool2"
  type: "Pooling"
  bottom: "norm2"
  top: "pool2"
  pooling_param {
    pool: MAX
    kernel_size: 3
    stride: 2
    pad: 0
  }
}
layer {
  name: "icp1_reduction1"
  type: "Convolution"
  bottom: "pool2"
  top: "icp1_reduction1"
  convolution_param {
    num_output: 96
    pad: 0
    kernel_size: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0.0
    }
  }
}
layer {
  name: "relu_icp1_reduction1"
  type: "ReLU"
  bottom: "icp1_reduction1"
  top: "icp1_reduction1"
}
layer {
  name: "icp1_reduction2"
  type: "Convolution"
  bottom: "pool2"
  top: "icp1_reduction2"
  convolution_param {
    num_output: 16
    pad: 0
    kernel_size: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0.0
    }
  }
}
layer {
  name: "relu_icp1_reduction2"
  type: "ReLU"
  bottom: "icp1_reduction2"
  top: "icp1_reduction2"
}
layer {
  name: "icp1_pool"
  type: "Pooling"
  bottom: "pool2"
  top: "icp1_pool"
  pooling_param {
    pool: MAX
    kernel_size: 3
    stride: 1
    pad: 1
  }
}
layer {
  name: "icp1_out0"
  type: "Convolution"
  bottom: "pool2"
  top: "icp1_out0"
  convolution_param {
    num_output: 64
    pad: 0
    kernel_size: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0.0
    }
  }
}
layer {
  name: "relu_icp1_out0"
  type: "ReLU"
  bottom: "icp1_out0"
  top: "icp1_out0"
}
layer {
  name: "icp1_out1"
  type: "Convolution"
  bottom: "icp1_reduction1"
  top: "icp1_out1"
  convolution_param {
    num_output: 128
    pad: 1
    kernel_size: 3
    weight_filler {
      type: "gaussian"
      std: 0.0399999991059
    }
    bias_filler {
      type: "constant"
      value: 0.0
    }
  }
}
layer {
  name: "relu_icp1_out1"
  type: "ReLU"
  bottom: "icp1_out1"
  top: "icp1_out1"
}
layer {
  name: "icp1_out2"
  type: "Convolution"
  bottom: "icp1_reduction2"
  top: "icp1_out2"
  convolution_param {
    num_output: 32
    pad: 2
    kernel_size: 5
    weight_filler {
      type: "gaussian"
      std: 0.0799999982119
    }
    bias_filler {
      type: "constant"
      value: 0.0
    }
  }
}
layer {
  name: "relu_icp1_out2"
  type: "ReLU"
  bottom: "icp1_out2"
  top: "icp1_out2"
}
layer {
  name: "icp1_out3"
  type: "Convolution"
  bottom: "icp1_pool"
  top: "icp1_out3"
  convolution_param {
    num_output: 32
    pad: 0
    kernel_size: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0.0
    }
  }
}
layer {
  name: "relu_icp1_out3"
  type: "ReLU"
  bottom: "icp1_out3"
  top: "icp1_out3"
}
layer {
  name: "icp2_in"
  type: "Concat"
  bottom: "icp1_out0"
  bottom: "icp1_out1"
  bottom: "icp1_out2"
  bottom: "icp1_out3"
  top: "icp2_in"
}
layer {
  name: "icp2_reduction1"
  type: "Convolution"
  bottom: "icp2_in"
  top: "icp2_reduction1"
  convolution_param {
    num_output: 128
    pad: 0
    kernel_size: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0.0
    }
  }
}
layer {
  name: "relu_icp2_reduction1"
  type: "ReLU"
  bottom: "icp2_reduction1"
  top: "icp2_reduction1"
}
layer {
  name: "icp2_reduction2"
  type: "Convolution"
  bottom: "icp2_in"
  top: "icp2_reduction2"
  convolution_param {
    num_output: 32
    pad: 0
    kernel_size: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0.0
    }
  }
}
layer {
  name: "relu_icp2_reduction2"
  type: "ReLU"
  bottom: "icp2_reduction2"
  top: "icp2_reduction2"
}
layer {
  name: "icp2_pool"
  type: "Pooling"
  bottom: "icp2_in"
  top: "icp2_pool"
  pooling_param {
    pool: MAX
    kernel_size: 3
    stride: 1
    pad: 1
  }
}
layer {
  name: "icp2_out0"
  type: "Convolution"
  bottom: "icp2_in"
  top: "icp2_out0"
  convolution_param {
    num_output: 128
    pad: 0
    kernel_size: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0.0
    }
  }
}
layer {
  name: "relu_icp2_out0"
  type: "ReLU"
  bottom: "icp2_out0"
  top: "icp2_out0"
}
layer {
  name: "icp2_out1"
  type: "Convolution"
  bottom: "icp2_reduction1"
  top: "icp2_out1"
  convolution_param {
    num_output: 192
    pad: 1
    kernel_size: 3
    weight_filler {
      type: "gaussian"
      std: 0.0399999991059
    }
    bias_filler {
      type: "constant"
      value: 0.0
    }
  }
}
layer {
  name: "relu_icp2_out1"
  type: "ReLU"
  bottom: "icp2_out1"
  top: "icp2_out1"
}
layer {
  name: "icp2_out2"
  type: "Convolution"
  bottom: "icp2_reduction2"
  top: "icp2_out2"
  convolution_param {
    num_output: 96
    pad: 2
    kernel_size: 5
    weight_filler {
      type: "gaussian"
      std: 0.0799999982119
    }
    bias_filler {
      type: "constant"
      value: 0.0
    }
  }
}
layer {
  name: "relu_icp2_out2"
  type: "ReLU"
  bottom: "icp2_out2"
  top: "icp2_out2"
}
layer {
  name: "icp2_out3"
  type: "Convolution"
  bottom: "icp2_pool"
  top: "icp2_out3"
  convolution_param {
    num_output: 64
    pad: 0
    kernel_size: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0.0
    }
  }
}
layer {
  name: "relu_icp2_out3"
  type: "ReLU"
  bottom: "icp2_out3"
  top: "icp2_out3"
}
layer {
  name: "icp2_out"
  type: "Concat"
  bottom: "icp2_out0"
  bottom: "icp2_out1"
  bottom: "icp2_out2"
  bottom: "icp2_out3"
  top: "icp2_out"
}
layer {
  name: "icp3_in"
  type: "Pooling"
  bottom: "icp2_out"
  top: "icp3_in"
  pooling_param {
    pool: MAX
    kernel_size: 3
    stride: 2
    pad: 0
  }
}
layer {
  name: "icp3_reduction1"
  type: "Convolution"
  bottom: "icp3_in"
  top: "icp3_reduction1"
  convolution_param {
    num_output: 96
    pad: 0
    kernel_size: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0.0
    }
  }
}
layer {
  name: "relu_icp3_reduction1"
  type: "ReLU"
  bottom: "icp3_reduction1"
  top: "icp3_reduction1"
}
layer {
  name: "icp3_reduction2"
  type: "Convolution"
  bottom: "icp3_in"
  top: "icp3_reduction2"
  convolution_param {
    num_output: 16
    pad: 0
    kernel_size: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0.0
    }
  }
}
layer {
  name: "relu_icp3_reduction2"
  type: "ReLU"
  bottom: "icp3_reduction2"
  top: "icp3_reduction2"
}
layer {
  name: "icp3_pool"
  type: "Pooling"
  bottom: "icp3_in"
  top: "icp3_pool"
  pooling_param {
    pool: MAX
    kernel_size: 3
    stride: 1
    pad: 1
  }
}
layer {
  name: "icp3_out0"
  type: "Convolution"
  bottom: "icp3_in"
  top: "icp3_out0"
  convolution_param {
    num_output: 192
    pad: 0
    kernel_size: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0.0
    }
  }
}
layer {
  name: "relu_icp3_out0"
  type: "ReLU"
  bottom: "icp3_out0"
  top: "icp3_out0"
}
layer {
  name: "icp3_out1"
  type: "Convolution"
  bottom: "icp3_reduction1"
  top: "icp3_out1"
  convolution_param {
    num_output: 208
    pad: 1
    kernel_size: 3
    weight_filler {
      type: "gaussian"
      std: 0.0399999991059
    }
    bias_filler {
      type: "constant"
      value: 0.0
    }
  }
}
layer {
  name: "relu_icp3_out1"
  type: "ReLU"
  bottom: "icp3_out1"
  top: "icp3_out1"
}
layer {
  name: "icp3_out2"
  type: "Convolution"
  bottom: "icp3_reduction2"
  top: "icp3_out2"
  convolution_param {
    num_output: 48
    pad: 2
    kernel_size: 5
    weight_filler {
      type: "gaussian"
      std: 0.0799999982119
    }
    bias_filler {
      type: "constant"
      value: 0.0
    }
  }
}
layer {
  name: "relu_icp3_out2"
  type: "ReLU"
  bottom: "icp3_out2"
  top: "icp3_out2"
}
layer {
  name: "icp3_out3"
  type: "Convolution"
  bottom: "icp3_pool"
  top: "icp3_out3"
  convolution_param {
    num_output: 64
    pad: 0
    kernel_size: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0.0
    }
  }
}
layer {
  name: "relu_icp3_out3"
  type: "ReLU"
  bottom: "icp3_out3"
  top: "icp3_out3"
}
layer {
  name: "icp4_in"
  type: "Concat"
  bottom: "icp3_out0"
  bottom: "icp3_out1"
  bottom: "icp3_out2"
  bottom: "icp3_out3"
  top: "icp4_in"
}
layer {
  name: "lstm1"
  type: "LSTM"
  bottom: "icp4_in"
  bottom: "clip"
  top: "lstm1"
  recurrent_param {
    num_output: 1000
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0.0
    }
  }
}
layer {
  name: "lstm2"
  type: "LSTM"
  bottom: "lstm1"
  bottom: "clip"
  top: "lstm2"
  recurrent_param {
    num_output: 1000
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0.0
    }
  }
}
layer {
  name: "reg_pool"
  type: "Pooling"
  bottom: "lstm2"
  top: "reg_pool"
  pooling_param {
    pool: AVE
    kernel_size: 7
    stride: 1
    pad: 0
  }
}
layer {
  name: "reg_fc1"
  type: "InnerProduct"
  bottom: "reg_pool"
  top: "reg_fc1"
  inner_product_param {
    num_output: 2048
    weight_filler {
      type: "gaussian"
      std: 0.00999999977648
    }
    bias_filler {
      type: "constant"
      value: 0.0
    }
  }
}
layer {
  name: "reg_relu"
  type: "ReLU"
  bottom: "reg_fc1"
  top: "reg_fc1"
}
layer {
  name: "reg_dropout"
  type: "Dropout"
  bottom: "reg_fc1"
  top: "reg_dropout"
  dropout_param {
    dropout_ratio: 0.5
  }
}
layer {
  name: "reg_fc2_xyz"
  type: "InnerProduct"
  bottom: "reg_dropout"
  top: "reg_fc2_xyz"
  inner_product_param {
    num_output: 3
    weight_filler {
      type: "gaussian"
      std: 0.5
    }
    bias_filler {
      type: "constant"
      value: 0.0
    }
  }
}
layer {
  name: "reg_loss_xyz"
  type: "EuclideanLoss"
  bottom: "reg_fc2_xyz"
  bottom: "label_xyz"
  top: "reg_loss_xyz"
}
layer {
  name: "reg_fc2_wpqr"
  type: "InnerProduct"
  bottom: "reg_dropout"
  top: "reg_fc2_wpqr"
  inner_product_param {
    num_output: 4
    weight_filler {
      type: "gaussian"
      std: 0.5
    }
    bias_filler {
      type: "constant"
      value: 0.0
    }
  }
}
layer {
  name: "reg_loss_wpqr"
  type: "EuclideanLoss"
  bottom: "reg_fc2_wpqr"
  bottom: "label_wpqr"
  top: "reg_loss_wpqr"
}

and this is the caffe output:

I0809 13:12:25.560395  3624 solver.cpp:44] Initializing solver from parameters: 
train_net: "deep6DoF_auto_train_net.prototxt"
test_net: "deep6DoF_auto_test_net.prototxt"
test_iter: 50
test_interval: 100
base_lr: 0.001
display: 50
max_iter: 500
lr_policy: "step"
gamma: 0.1
momentum: 0.9
weight_decay: 0.5
stepsize: 150
snapshot: 5000
snapshot_prefix: "snapshots/train"
delta: 1e-08
test_initialization: true
momentum2: 0.999
I0809 13:12:25.562652  3624 solver.cpp:77] Creating training net from train_net file: deep6DoF_auto_train_net.prototxt
I0809 13:12:25.564133  3624 upgrade_proto.cpp:67] Attempting to upgrade input file specified using deprecated input fields: deep6DoF_auto_train_net.prototxt
I0809 13:12:25.564175  3624 upgrade_proto.cpp:70] Successfully upgraded file specified using deprecated input fields.
W0809 13:12:25.564185  3624 upgrade_proto.cpp:72] Note that future Caffe releases will only support input layers and not input fields.
I0809 13:12:25.564687  3624 net.cpp:51] Initializing net from parameters: 
state {
  phase: TRAIN
}
layer {
  name: "input"
  type: "Input"
  top: "clip"
  input_param {
    shape {
      dim: 1
      dim: 51
    }
  }
}
layer {
  name: "data"
  type: "Data"
  top: "data"
  data_param {
    source: "/home/yasin/lmdb_dataset/train_lmdb"
    batch_size: 1
    backend: LMDB
  }
}
layer {
  name: "label"
  type: "Data"
  top: "label"
  data_param {
    source: "/home/yasin/lmdb_dataset/train_label_lmdb"
    batch_size: 1
    backend: LMDB
  }
}
layer {
  name: "label_wpqr"
  type: "Slice"
  bottom: "label"
  top: "label_wpqr"
  top: "label_xyz"
  slice_param {
    slice_point: 4
    axis: 1
  }
}
layer {
  name: "conv1"
  type: "Convolution"
  bottom: "data"
  top: "conv1"
  convolution_param {
    num_output: 64
    pad: 3
    kernel_size: 7
    stride: 2
    weight_filler {
      type: "gaussian"
      std: 0.015
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
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
    pad: 0
  }
}
layer {
  name: "norm1"
  type: "LRN"
  bottom: "pool1"
  top: "norm1"
  lrn_param {
    local_size: 5
    alpha: 0.0001
    beta: 0.75
  }
}
layer {
  name: "reduction2"
  type: "Convolution"
  bottom: "norm1"
  top: "reduction2"
  convolution_param {
    num_output: 64
    pad: 0
    kernel_size: 1
    group: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "relu_reduction2"
  type: "ReLU"
  bottom: "reduction2"
  top: "reduction2"
}
layer {
  name: "conv2"
  type: "Convolution"
  bottom: "conv1"
  top: "conv2"
  convolution_param {
    num_output: 192
    pad: 1
    kernel_size: 3
    group: 1
    weight_filler {
      type: "gaussian"
      std: 0.02
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
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
    pad: 0
  }
}
layer {
  name: "icp1_reduction1"
  type: "Convolution"
  bottom: "pool2"
  top: "icp1_reduction1"
  convolution_param {
    num_output: 96
    pad: 0
    kernel_size: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "relu_icp1_reduction1"
  type: "ReLU"
  bottom: "icp1_reduction1"
  top: "icp1_reduction1"
}
layer {
  name: "icp1_reduction2"
  type: "Convolution"
  bottom: "pool2"
  top: "icp1_reduction2"
  convolution_param {
    num_output: 16
    pad: 0
    kernel_size: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "relu_icp1_reduction2"
  type: "ReLU"
  bottom: "icp1_reduction2"
  top: "icp1_reduction2"
}
layer {
  name: "icp1_pool"
  type: "Pooling"
  bottom: "pool2"
  top: "icp1_pool"
  pooling_param {
    pool: MAX
    kernel_size: 3
    stride: 1
    pad: 1
  }
}
layer {
  name: "icp1_out0"
  type: "Convolution"
  bottom: "pool2"
  top: "icp1_out0"
  convolution_param {
    num_output: 64
    pad: 0
    kernel_size: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "relu_icp1_out0"
  type: "ReLU"
  bottom: "icp1_out0"
  top: "icp1_out0"
}
layer {
  name: "icp1_out1"
  type: "Convolution"
  bottom: "icp1_reduction1"
  top: "icp1_out1"
  convolution_param {
    num_output: 128
    pad: 1
    kernel_size: 3
    weight_filler {
      type: "gaussian"
      std: 0.04
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "relu_icp1_out1"
  type: "ReLU"
  bottom: "icp1_out1"
  top: "icp1_out1"
}
layer {
  name: "icp1_out2"
  type: "Convolution"
  bottom: "icp1_reduction2"
  top: "icp1_out2"
  convolution_param {
    num_output: 32
    pad: 2
    kernel_size: 5
    weight_filler {
      type: "gaussian"
      std: 0.08
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "relu_icp1_out2"
  type: "ReLU"
  bottom: "icp1_out2"
  top: "icp1_out2"
}
layer {
  name: "icp1_out3"
  type: "Convolution"
  bottom: "icp1_pool"
  top: "icp1_out3"
  convolution_param {
    num_output: 32
    pad: 0
    kernel_size: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "relu_icp1_out3"
  type: "ReLU"
  bottom: "icp1_out3"
  top: "icp1_out3"
}
layer {
  name: "icp2_in"
  type: "Concat"
  bottom: "icp1_out0"
  bottom: "icp1_out1"
  bottom: "icp1_out2"
  bottom: "icp1_out3"
  top: "icp2_in"
}
layer {
  name: "icp2_reduction1"
  type: "Convolution"
  bottom: "icp2_in"
  top: "icp2_reduction1"
  convolution_param {
    num_output: 128
    pad: 0
    kernel_size: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "relu_icp2_reduction1"
  type: "ReLU"
  bottom: "icp2_reduction1"
  top: "icp2_reduction1"
}
layer {
  name: "icp2_reduction2"
  type: "Convolution"
  bottom: "icp2_in"
  top: "icp2_reduction2"
  convolution_param {
    num_output: 32
    pad: 0
    kernel_size: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "relu_icp2_reduction2"
  type: "ReLU"
  bottom: "icp2_reduction2"
  top: "icp2_reduction2"
}
layer {
  name: "icp2_pool"
  type: "Pooling"
  bottom: "icp2_in"
  top: "icp2_pool"
  pooling_param {
    pool: MAX
    kernel_size: 3
    stride: 1
    pad: 1
  }
}
layer {
  name: "icp2_out0"
  type: "Convolution"
  bottom: "icp2_in"
  top: "icp2_out0"
  convolution_param {
    num_output: 128
    pad: 0
    kernel_size: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "relu_icp2_out0"
  type: "ReLU"
  bottom: "icp2_out0"
  top: "icp2_out0"
}
layer {
  name: "icp2_out1"
  type: "Convolution"
  bottom: "icp2_reduction1"
  top: "icp2_out1"
  convolution_param {
    num_output: 192
    pad: 1
    kernel_size: 3
    weight_filler {
      type: "gaussian"
      std: 0.04
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "relu_icp2_out1"
  type: "ReLU"
  bottom: "icp2_out1"
  top: "icp2_out1"
}
layer {
  name: "icp2_out2"
  type: "Convolution"
  bottom: "icp2_reduction2"
  top: "icp2_out2"
  convolution_param {
    num_output: 96
    pad: 2
    kernel_size: 5
    weight_filler {
      type: "gaussian"
      std: 0.08
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "relu_icp2_out2"
  type: "ReLU"
  bottom: "icp2_out2"
  top: "icp2_out2"
}
layer {
  name: "icp2_out3"
  type: "Convolution"
  bottom: "icp2_pool"
  top: "icp2_out3"
  convolution_param {
    num_output: 64
    pad: 0
    kernel_size: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "relu_icp2_out3"
  type: "ReLU"
  bottom: "icp2_out3"
  top: "icp2_out3"
}
layer {
  name: "icp2_out"
  type: "Concat"
  bottom: "icp2_out0"
  bottom: "icp2_out1"
  bottom: "icp2_out2"
  bottom: "icp2_out3"
  top: "icp2_out"
}
layer {
  name: "icp3_in"
  type: "Pooling"
  bottom: "icp2_out"
  top: "icp3_in"
  pooling_param {
    pool: MAX
    kernel_size: 3
    stride: 2
    pad: 0
  }
}
layer {
  name: "icp3_reduction1"
  type: "Convolution"
  bottom: "icp3_in"
  top: "icp3_reduction1"
  convolution_param {
    num_output: 96
    pad: 0
    kernel_size: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "relu_icp3_reduction1"
  type: "ReLU"
  bottom: "icp3_reduction1"
  top: "icp3_reduction1"
}
layer {
  name: "icp3_reduction2"
  type: "Convolution"
  bottom: "icp3_in"
  top: "icp3_reduction2"
  convolution_param {
    num_output: 16
    pad: 0
    kernel_size: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "relu_icp3_reduction2"
  type: "ReLU"
  bottom: "icp3_reduction2"
  top: "icp3_reduction2"
}
layer {
  name: "icp3_pool"
  type: "Pooling"
  bottom: "icp3_in"
  top: "icp3_pool"
  pooling_param {
    pool: MAX
    kernel_size: 3
    stride: 1
    pad: 1
  }
}
layer {
  name: "icp3_out0"
  type: "Convolution"
  bottom: "icp3_in"
  top: "icp3_out0"
  convolution_param {
    num_output: 192
    pad: 0
    kernel_size: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "relu_icp3_out0"
  type: "ReLU"
  bottom: "icp3_out0"
  top: "icp3_out0"
}
layer {
  name: "icp3_out1"
  type: "Convolution"
  bottom: "icp3_reduction1"
  top: "icp3_out1"
  convolution_param {
    num_output: 208
    pad: 1
    kernel_size: 3
    weight_filler {
      type: "gaussian"
      std: 0.04
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "relu_icp3_out1"
  type: "ReLU"
  bottom: "icp3_out1"
  top: "icp3_out1"
}
layer {
  name: "icp3_out2"
  type: "Convolution"
  bottom: "icp3_reduction2"
  top: "icp3_out2"
  convolution_param {
    num_output: 48
    pad: 2
    kernel_size: 5
    weight_filler {
      type: "gaussian"
      std: 0.08
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "relu_icp3_out2"
  type: "ReLU"
  bottom: "icp3_out2"
  top: "icp3_out2"
}
layer {
  name: "icp3_out3"
  type: "Convolution"
  bottom: "icp3_pool"
  top: "icp3_out3"
  convolution_param {
    num_output: 64
    pad: 0
    kernel_size: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "relu_icp3_out3"
  type: "ReLU"
  bottom: "icp3_out3"
  top: "icp3_out3"
}
layer {
  name: "icp4_in"
  type: "Concat"
  bottom: "icp3_out0"
  bottom: "icp3_out1"
  bottom: "icp3_out2"
  bottom: "icp3_out3"
  top: "icp4_in"
}
layer {
  name: "lstm1"
  type: "LSTM"
  bottom: "icp4_in"
  bottom: "clip"
  top: "lstm1"
  recurrent_param {
    num_output: 1000
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "lstm2"
  type: "LSTM"
  bottom: "lstm1"
  bottom: "clip"
  top: "lstm2"
  recurrent_param {
    num_output: 1000
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "reg_pool"
  type: "Pooling"
  bottom: "lstm2"
  top: "reg_pool"
  pooling_param {
    pool: AVE
    kernel_size: 7
    stride: 1
    pad: 0
  }
}
layer {
  name: "reg_fc1"
  type: "InnerProduct"
  bottom: "reg_pool"
  top: "reg_fc1"
  inner_product_param {
    num_output: 2048
    weight_filler {
      type: "gaussian"
      std: 0.01
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "reg_relu"
  type: "ReLU"
  bottom: "reg_fc1"
  top: "reg_fc1"
}
layer {
  name: "reg_dropout"
  type: "Dropout"
  bottom: "reg_fc1"
  top: "reg_dropout"
  dropout_param {
    dropout_ratio: 0.5
  }
}
layer {
  name: "reg_fc2_xyz"
  type: "InnerProduct"
  bottom: "reg_dropout"
  top: "reg_fc2_xyz"
  inner_product_param {
    num_output: 3
    weight_filler {
      type: "gaussian"
      std: 0.5
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "reg_loss_xyz"
  type: "EuclideanLoss"
  bottom: "reg_fc2_xyz"
  bottom: "label_xyz"
  top: "reg_loss_xyz"
}
layer {
  name: "reg_fc2_wpqr"
  type: "InnerProduct"
  bottom: "reg_dropout"
  top: "reg_fc2_wpqr"
  inner_product_param {
    num_output: 4
    weight_filler {
      type: "gaussian"
      std: 0.5
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "reg_loss_wpqr"
  type: "EuclideanLoss"
  bottom: "reg_fc2_wpqr"
  bottom: "label_wpqr"
  top: "reg_loss_wpqr"
}
I0809 13:12:25.565456  3624 layer_factory.hpp:77] Creating layer input
I0809 13:12:25.565492  3624 net.cpp:84] Creating Layer input
I0809 13:12:25.565503  3624 net.cpp:380] input -> clip
I0809 13:12:25.565529  3624 net.cpp:122] Setting up input
I0809 13:12:25.565573  3624 net.cpp:129] Top shape: 1 51 (51)
I0809 13:12:25.565582  3624 net.cpp:137] Memory required for data: 204
I0809 13:12:25.565590  3624 layer_factory.hpp:77] Creating layer clip_input_0_split
I0809 13:12:25.565620  3624 net.cpp:84] Creating Layer clip_input_0_split
I0809 13:12:25.565647  3624 net.cpp:406] clip_input_0_split <- clip
I0809 13:12:25.565660  3624 net.cpp:380] clip_input_0_split -> clip_input_0_split_0
I0809 13:12:25.565673  3624 net.cpp:380] clip_input_0_split -> clip_input_0_split_1
I0809 13:12:25.565690  3624 net.cpp:122] Setting up clip_input_0_split
I0809 13:12:25.565702  3624 net.cpp:129] Top shape: 1 51 (51)
I0809 13:12:25.565728  3624 net.cpp:129] Top shape: 1 51 (51)
I0809 13:12:25.565737  3624 net.cpp:137] Memory required for data: 612
I0809 13:12:25.565743  3624 layer_factory.hpp:77] Creating layer data
I0809 13:12:25.565876  3624 db_lmdb.cpp:35] Opened lmdb /home/yasin/lmdb_dataset/train_lmdb
I0809 13:12:25.565922  3624 net.cpp:84] Creating Layer data
I0809 13:12:25.565939  3624 net.cpp:380] data -> data
I0809 13:12:25.567041  3624 data_layer.cpp:45] output data size: 1,4,480,640
I0809 13:12:25.567515  3624 net.cpp:122] Setting up data
I0809 13:12:25.567546  3624 net.cpp:129] Top shape: 1 4 480 640 (1228800)
I0809 13:12:25.567555  3624 net.cpp:137] Memory required for data: 4915812
I0809 13:12:25.567579  3624 layer_factory.hpp:77] Creating layer label
I0809 13:12:25.568075  3624 db_lmdb.cpp:35] Opened lmdb /home/yasin/lmdb_dataset/train_label_lmdb
I0809 13:12:25.568128  3624 net.cpp:84] Creating Layer label
I0809 13:12:25.568142  3624 net.cpp:380] label -> label
I0809 13:12:25.569280  3624 data_layer.cpp:45] output data size: 1,4,480,640
I0809 13:12:25.570099  3624 net.cpp:122] Setting up label
I0809 13:12:25.570135  3624 net.cpp:129] Top shape: 1 4 480 640 (1228800)
I0809 13:12:25.570144  3624 net.cpp:137] Memory required for data: 9831012
I0809 13:12:25.570154  3624 layer_factory.hpp:77] Creating layer label_wpqr
I0809 13:12:25.570173  3624 net.cpp:84] Creating Layer label_wpqr
I0809 13:12:25.570220  3624 net.cpp:406] label_wpqr <- label
I0809 13:12:25.570236  3624 net.cpp:380] label_wpqr -> label_wpqr
I0809 13:12:25.570264  3624 net.cpp:380] label_wpqr -> label_xyz
I0809 13:12:25.570294  3624 net.cpp:122] Setting up label_wpqr
I0809 13:12:25.570323  3624 net.cpp:129] Top shape: 1 4 480 640 (1228800)
I0809 13:12:25.570333  3624 net.cpp:129] Top shape: 1 0 480 640 (0)
I0809 13:12:25.570340  3624 net.cpp:137] Memory required for data: 14746212
I0809 13:12:25.570348  3624 layer_factory.hpp:77] Creating layer conv1
I0809 13:12:25.570369  3624 net.cpp:84] Creating Layer conv1
I0809 13:12:25.570392  3624 net.cpp:406] conv1 <- data
I0809 13:12:25.570405  3624 net.cpp:380] conv1 -> conv1
I0809 13:12:25.570736  3624 net.cpp:122] Setting up conv1
I0809 13:12:25.570772  3624 net.cpp:129] Top shape: 1 64 240 320 (4915200)
I0809 13:12:25.570780  3624 net.cpp:137] Memory required for data: 34407012
I0809 13:12:25.570797  3624 layer_factory.hpp:77] Creating layer relu1
I0809 13:12:25.570809  3624 net.cpp:84] Creating Layer relu1
I0809 13:12:25.570821  3624 net.cpp:406] relu1 <- conv1
I0809 13:12:25.570833  3624 net.cpp:367] relu1 -> conv1 (in-place)
I0809 13:12:25.570861  3624 net.cpp:122] Setting up relu1
I0809 13:12:25.570873  3624 net.cpp:129] Top shape: 1 64 240 320 (4915200)
I0809 13:12:25.570896  3624 net.cpp:137] Memory required for data: 54067812
I0809 13:12:25.570921  3624 layer_factory.hpp:77] Creating layer conv1_relu1_0_split
I0809 13:12:25.570952  3624 net.cpp:84] Creating Layer conv1_relu1_0_split
I0809 13:12:25.570961  3624 net.cpp:406] conv1_relu1_0_split <- conv1
I0809 13:12:25.570987  3624 net.cpp:380] conv1_relu1_0_split -> conv1_relu1_0_split_0
I0809 13:12:25.571005  3624 net.cpp:380] conv1_relu1_0_split -> conv1_relu1_0_split_1
I0809 13:12:25.571019  3624 net.cpp:122] Setting up conv1_relu1_0_split
I0809 13:12:25.571030  3624 net.cpp:129] Top shape: 1 64 240 320 (4915200)
I0809 13:12:25.571056  3624 net.cpp:129] Top shape: 1 64 240 320 (4915200)
I0809 13:12:25.571064  3624 net.cpp:137] Memory required for data: 93389412
I0809 13:12:25.571071  3624 layer_factory.hpp:77] Creating layer pool1
I0809 13:12:25.571087  3624 net.cpp:84] Creating Layer pool1
I0809 13:12:25.571099  3624 net.cpp:406] pool1 <- conv1_relu1_0_split_0
I0809 13:12:25.571130  3624 net.cpp:380] pool1 -> pool1
I0809 13:12:25.571156  3624 net.cpp:122] Setting up pool1
I0809 13:12:25.571183  3624 net.cpp:129] Top shape: 1 64 120 160 (1228800)
I0809 13:12:25.571190  3624 net.cpp:137] Memory required for data: 98304612
I0809 13:12:25.571198  3624 layer_factory.hpp:77] Creating layer norm1
I0809 13:12:25.571214  3624 net.cpp:84] Creating Layer norm1
I0809 13:12:25.571223  3624 net.cpp:406] norm1 <- pool1
I0809 13:12:25.571250  3624 net.cpp:380] norm1 -> norm1
I0809 13:12:25.571264  3624 net.cpp:122] Setting up norm1
I0809 13:12:25.571275  3624 net.cpp:129] Top shape: 1 64 120 160 (1228800)
I0809 13:12:25.571282  3624 net.cpp:137] Memory required for data: 103219812
I0809 13:12:25.571290  3624 layer_factory.hpp:77] Creating layer reduction2
I0809 13:12:25.571321  3624 net.cpp:84] Creating Layer reduction2
I0809 13:12:25.571331  3624 net.cpp:406] reduction2 <- norm1
I0809 13:12:25.571344  3624 net.cpp:380] reduction2 -> reduction2
I0809 13:12:25.571439  3624 net.cpp:122] Setting up reduction2
I0809 13:12:25.571487  3624 net.cpp:129] Top shape: 1 64 120 160 (1228800)
I0809 13:12:25.571497  3624 net.cpp:137] Memory required for data: 108135012
I0809 13:12:25.571512  3624 layer_factory.hpp:77] Creating layer relu_reduction2
I0809 13:12:25.571527  3624 net.cpp:84] Creating Layer relu_reduction2
I0809 13:12:25.571550  3624 net.cpp:406] relu_reduction2 <- reduction2
I0809 13:12:25.571561  3624 net.cpp:367] relu_reduction2 -> reduction2 (in-place)
I0809 13:12:25.571573  3624 net.cpp:122] Setting up relu_reduction2
I0809 13:12:25.571583  3624 net.cpp:129] Top shape: 1 64 120 160 (1228800)
I0809 13:12:25.571596  3624 net.cpp:137] Memory required for data: 113050212
I0809 13:12:25.571606  3624 layer_factory.hpp:77] Creating layer conv2
I0809 13:12:25.571624  3624 net.cpp:84] Creating Layer conv2
I0809 13:12:25.571648  3624 net.cpp:406] conv2 <- conv1_relu1_0_split_1
I0809 13:12:25.571661  3624 net.cpp:380] conv2 -> conv2
I0809 13:12:25.573698  3624 net.cpp:122] Setting up conv2
I0809 13:12:25.573761  3624 net.cpp:129] Top shape: 1 192 240 320 (14745600)
I0809 13:12:25.573770  3624 net.cpp:137] Memory required for data: 172032612
I0809 13:12:25.573791  3624 layer_factory.hpp:77] Creating layer relu2
I0809 13:12:25.573807  3624 net.cpp:84] Creating Layer relu2
I0809 13:12:25.573817  3624 net.cpp:406] relu2 <- conv2
I0809 13:12:25.573829  3624 net.cpp:367] relu2 -> conv2 (in-place)
I0809 13:12:25.573922  3624 net.cpp:122] Setting up relu2
I0809 13:12:25.573940  3624 net.cpp:129] Top shape: 1 192 240 320 (14745600)
I0809 13:12:25.573947  3624 net.cpp:137] Memory required for data: 231015012
I0809 13:12:25.573956  3624 layer_factory.hpp:77] Creating layer norm2
I0809 13:12:25.573968  3624 net.cpp:84] Creating Layer norm2
I0809 13:12:25.573977  3624 net.cpp:406] norm2 <- conv2
I0809 13:12:25.573990  3624 net.cpp:380] norm2 -> norm2
I0809 13:12:25.574005  3624 net.cpp:122] Setting up norm2
I0809 13:12:25.574015  3624 net.cpp:129] Top shape: 1 192 240 320 (14745600)
I0809 13:12:25.574023  3624 net.cpp:137] Memory required for data: 289997412
I0809 13:12:25.574030  3624 layer_factory.hpp:77] Creating layer pool2
I0809 13:12:25.574043  3624 net.cpp:84] Creating Layer pool2
I0809 13:12:25.574050  3624 net.cpp:406] pool2 <- norm2
I0809 13:12:25.574060  3624 net.cpp:380] pool2 -> pool2
I0809 13:12:25.574075  3624 net.cpp:122] Setting up pool2
I0809 13:12:25.574084  3624 net.cpp:129] Top shape: 1 192 120 160 (3686400)
I0809 13:12:25.574091  3624 net.cpp:137] Memory required for data: 304743012
I0809 13:12:25.574100  3624 layer_factory.hpp:77] Creating layer pool2_pool2_0_split
I0809 13:12:25.574116  3624 net.cpp:84] Creating Layer pool2_pool2_0_split
I0809 13:12:25.574124  3624 net.cpp:406] pool2_pool2_0_split <- pool2
I0809 13:12:25.574134  3624 net.cpp:380] pool2_pool2_0_split -> pool2_pool2_0_split_0
I0809 13:12:25.574151  3624 net.cpp:380] pool2_pool2_0_split -> pool2_pool2_0_split_1
I0809 13:12:25.574164  3624 net.cpp:380] pool2_pool2_0_split -> pool2_pool2_0_split_2
I0809 13:12:25.574177  3624 net.cpp:380] pool2_pool2_0_split -> pool2_pool2_0_split_3
I0809 13:12:25.574190  3624 net.cpp:122] Setting up pool2_pool2_0_split
I0809 13:12:25.574200  3624 net.cpp:129] Top shape: 1 192 120 160 (3686400)
I0809 13:12:25.574209  3624 net.cpp:129] Top shape: 1 192 120 160 (3686400)
I0809 13:12:25.574218  3624 net.cpp:129] Top shape: 1 192 120 160 (3686400)
I0809 13:12:25.574226  3624 net.cpp:129] Top shape: 1 192 120 160 (3686400)
I0809 13:12:25.574234  3624 net.cpp:137] Memory required for data: 363725412
I0809 13:12:25.574241  3624 layer_factory.hpp:77] Creating layer icp1_reduction1
I0809 13:12:25.574259  3624 net.cpp:84] Creating Layer icp1_reduction1
I0809 13:12:25.574267  3624 net.cpp:406] icp1_reduction1 <- pool2_pool2_0_split_0
I0809 13:12:25.574280  3624 net.cpp:380] icp1_reduction1 -> icp1_reduction1
I0809 13:12:25.574645  3624 net.cpp:122] Setting up icp1_reduction1
I0809 13:12:25.574681  3624 net.cpp:129] Top shape: 1 96 120 160 (1843200)
I0809 13:12:25.574689  3624 net.cpp:137] Memory required for data: 371098212
I0809 13:12:25.574702  3624 layer_factory.hpp:77] Creating layer relu_icp1_reduction1
I0809 13:12:25.574729  3624 net.cpp:84] Creating Layer relu_icp1_reduction1
I0809 13:12:25.574739  3624 net.cpp:406] relu_icp1_reduction1 <- icp1_reduction1
I0809 13:12:25.574767  3624 net.cpp:367] relu_icp1_reduction1 -> icp1_reduction1 (in-place)
I0809 13:12:25.574782  3624 net.cpp:122] Setting up relu_icp1_reduction1
I0809 13:12:25.574806  3624 net.cpp:129] Top shape: 1 96 120 160 (1843200)
I0809 13:12:25.574815  3624 net.cpp:137] Memory required for data: 378471012
I0809 13:12:25.574822  3624 layer_factory.hpp:77] Creating layer icp1_reduction2
I0809 13:12:25.574851  3624 net.cpp:84] Creating Layer icp1_reduction2
I0809 13:12:25.574861  3624 net.cpp:406] icp1_reduction2 <- pool2_pool2_0_split_1
I0809 13:12:25.574887  3624 net.cpp:380] icp1_reduction2 -> icp1_reduction2
I0809 13:12:25.574972  3624 net.cpp:122] Setting up icp1_reduction2
I0809 13:12:25.575001  3624 net.cpp:129] Top shape: 1 16 120 160 (307200)
I0809 13:12:25.575009  3624 net.cpp:137] Memory required for data: 379699812
I0809 13:12:25.575026  3624 layer_factory.hpp:77] Creating layer relu_icp1_reduction2
I0809 13:12:25.575037  3624 net.cpp:84] Creating Layer relu_icp1_reduction2
I0809 13:12:25.575047  3624 net.cpp:406] relu_icp1_reduction2 <- icp1_reduction2
I0809 13:12:25.575057  3624 net.cpp:367] relu_icp1_reduction2 -> icp1_reduction2 (in-place)
I0809 13:12:25.575067  3624 net.cpp:122] Setting up relu_icp1_reduction2
I0809 13:12:25.575076  3624 net.cpp:129] Top shape: 1 16 120 160 (307200)
I0809 13:12:25.575083  3624 net.cpp:137] Memory required for data: 380928612
I0809 13:12:25.575090  3624 layer_factory.hpp:77] Creating layer icp1_pool
I0809 13:12:25.575103  3624 net.cpp:84] Creating Layer icp1_pool
I0809 13:12:25.575111  3624 net.cpp:406] icp1_pool <- pool2_pool2_0_split_2
I0809 13:12:25.575121  3624 net.cpp:380] icp1_pool -> icp1_pool
I0809 13:12:25.575136  3624 net.cpp:122] Setting up icp1_pool
I0809 13:12:25.575146  3624 net.cpp:129] Top shape: 1 192 120 160 (3686400)
I0809 13:12:25.575153  3624 net.cpp:137] Memory required for data: 395674212
I0809 13:12:25.575160  3624 layer_factory.hpp:77] Creating layer icp1_out0
I0809 13:12:25.575175  3624 net.cpp:84] Creating Layer icp1_out0
I0809 13:12:25.575183  3624 net.cpp:406] icp1_out0 <- pool2_pool2_0_split_3
I0809 13:12:25.575199  3624 net.cpp:380] icp1_out0 -> icp1_out0
I0809 13:12:25.575389  3624 net.cpp:122] Setting up icp1_out0
I0809 13:12:25.575428  3624 net.cpp:129] Top shape: 1 64 120 160 (1228800)
I0809 13:12:25.575470  3624 net.cpp:137] Memory required for data: 400589412
I0809 13:12:25.575484  3624 layer_factory.hpp:77] Creating layer relu_icp1_out0
I0809 13:12:25.575496  3624 net.cpp:84] Creating Layer relu_icp1_out0
I0809 13:12:25.575521  3624 net.cpp:406] relu_icp1_out0 <- icp1_out0
I0809 13:12:25.575538  3624 net.cpp:367] relu_icp1_out0 -> icp1_out0 (in-place)
I0809 13:12:25.575567  3624 net.cpp:122] Setting up relu_icp1_out0
I0809 13:12:25.575578  3624 net.cpp:129] Top shape: 1 64 120 160 (1228800)
I0809 13:12:25.575600  3624 net.cpp:137] Memory required for data: 405504612
I0809 13:12:25.575609  3624 layer_factory.hpp:77] Creating layer icp1_out1
I0809 13:12:25.575626  3624 net.cpp:84] Creating Layer icp1_out1
I0809 13:12:25.575650  3624 net.cpp:406] icp1_out1 <- icp1_reduction1
I0809 13:12:25.575662  3624 net.cpp:380] icp1_out1 -> icp1_out1
I0809 13:12:25.577116  3624 net.cpp:122] Setting up icp1_out1
I0809 13:12:25.577173  3624 net.cpp:129] Top shape: 1 128 120 160 (2457600)
I0809 13:12:25.577181  3624 net.cpp:137] Memory required for data: 415335012
I0809 13:12:25.577227  3624 layer_factory.hpp:77] Creating layer relu_icp1_out1
I0809 13:12:25.577258  3624 net.cpp:84] Creating Layer relu_icp1_out1
I0809 13:12:25.577288  3624 net.cpp:406] relu_icp1_out1 <- icp1_out1
I0809 13:12:25.577307  3624 net.cpp:367] relu_icp1_out1 -> icp1_out1 (in-place)
I0809 13:12:25.577354  3624 net.cpp:122] Setting up relu_icp1_out1
I0809 13:12:25.577366  3624 net.cpp:129] Top shape: 1 128 120 160 (2457600)
I0809 13:12:25.577374  3624 net.cpp:137] Memory required for data: 425165412
I0809 13:12:25.577381  3624 layer_factory.hpp:77] Creating layer icp1_out2
I0809 13:12:25.577401  3624 net.cpp:84] Creating Layer icp1_out2
I0809 13:12:25.577430  3624 net.cpp:406] icp1_out2 <- icp1_reduction2
I0809 13:12:25.577457  3624 net.cpp:380] icp1_out2 -> icp1_out2
I0809 13:12:25.578202  3624 net.cpp:122] Setting up icp1_out2
I0809 13:12:25.578274  3624 net.cpp:129] Top shape: 1 32 120 160 (614400)
I0809 13:12:25.578284  3624 net.cpp:137] Memory required for data: 427623012
I0809 13:12:25.578320  3624 layer_factory.hpp:77] Creating layer relu_icp1_out2
I0809 13:12:25.578344  3624 net.cpp:84] Creating Layer relu_icp1_out2
I0809 13:12:25.578374  3624 net.cpp:406] relu_icp1_out2 <- icp1_out2
I0809 13:12:25.578405  3624 net.cpp:367] relu_icp1_out2 -> icp1_out2 (in-place)
I0809 13:12:25.578439  3624 net.cpp:122] Setting up relu_icp1_out2
I0809 13:12:25.578465  3624 net.cpp:129] Top shape: 1 32 120 160 (614400)
I0809 13:12:25.578478  3624 net.cpp:137] Memory required for data: 430080612
I0809 13:12:25.578501  3624 layer_factory.hpp:77] Creating layer icp1_out3
I0809 13:12:25.578519  3624 net.cpp:84] Creating Layer icp1_out3
I0809 13:12:25.578543  3624 net.cpp:406] icp1_out3 <- icp1_pool
I0809 13:12:25.578557  3624 net.cpp:380] icp1_out3 -> icp1_out3
I0809 13:12:25.578737  3624 net.cpp:122] Setting up icp1_out3
I0809 13:12:25.578773  3624 net.cpp:129] Top shape: 1 32 120 160 (614400)
I0809 13:12:25.578780  3624 net.cpp:137] Memory required for data: 432538212
I0809 13:12:25.578835  3624 layer_factory.hpp:77] Creating layer relu_icp1_out3
I0809 13:12:25.578866  3624 net.cpp:84] Creating Layer relu_icp1_out3
I0809 13:12:25.578874  3624 net.cpp:406] relu_icp1_out3 <- icp1_out3
I0809 13:12:25.578902  3624 net.cpp:367] relu_icp1_out3 -> icp1_out3 (in-place)
I0809 13:12:25.578915  3624 net.cpp:122] Setting up relu_icp1_out3
I0809 13:12:25.578940  3624 net.cpp:129] Top shape: 1 32 120 160 (614400)
I0809 13:12:25.578948  3624 net.cpp:137] Memory required for data: 434995812
I0809 13:12:25.578955  3624 layer_factory.hpp:77] Creating layer icp2_in
I0809 13:12:25.578980  3624 net.cpp:84] Creating Layer icp2_in
I0809 13:12:25.578989  3624 net.cpp:406] icp2_in <- icp1_out0
I0809 13:12:25.579027  3624 net.cpp:406] icp2_in <- icp1_out1
I0809 13:12:25.579037  3624 net.cpp:406] icp2_in <- icp1_out2
I0809 13:12:25.579047  3624 net.cpp:406] icp2_in <- icp1_out3
I0809 13:12:25.579074  3624 net.cpp:380] icp2_in -> icp2_in
I0809 13:12:25.579109  3624 net.cpp:122] Setting up icp2_in
I0809 13:12:25.579138  3624 net.cpp:129] Top shape: 1 256 120 160 (4915200)
I0809 13:12:25.579164  3624 net.cpp:137] Memory required for data: 454656612
I0809 13:12:25.579186  3624 layer_factory.hpp:77] Creating layer icp2_in_icp2_in_0_split
I0809 13:12:25.579200  3624 net.cpp:84] Creating Layer icp2_in_icp2_in_0_split
I0809 13:12:25.579213  3624 net.cpp:406] icp2_in_icp2_in_0_split <- icp2_in
I0809 13:12:25.579231  3624 net.cpp:380] icp2_in_icp2_in_0_split -> icp2_in_icp2_in_0_split_0
I0809 13:12:25.579262  3624 net.cpp:380] icp2_in_icp2_in_0_split -> icp2_in_icp2_in_0_split_1
I0809 13:12:25.579277  3624 net.cpp:380] icp2_in_icp2_in_0_split -> icp2_in_icp2_in_0_split_2
I0809 13:12:25.579306  3624 net.cpp:380] icp2_in_icp2_in_0_split -> icp2_in_icp2_in_0_split_3
I0809 13:12:25.579334  3624 net.cpp:122] Setting up icp2_in_icp2_in_0_split
I0809 13:12:25.579365  3624 net.cpp:129] Top shape: 1 256 120 160 (4915200)
I0809 13:12:25.579375  3624 net.cpp:129] Top shape: 1 256 120 160 (4915200)
I0809 13:12:25.579383  3624 net.cpp:129] Top shape: 1 256 120 160 (4915200)
I0809 13:12:25.579392  3624 net.cpp:129] Top shape: 1 256 120 160 (4915200)
I0809 13:12:25.579419  3624 net.cpp:137] Memory required for data: 533299812
I0809 13:12:25.579432  3624 layer_factory.hpp:77] Creating layer icp2_reduction1
I0809 13:12:25.579479  3624 net.cpp:84] Creating Layer icp2_reduction1
I0809 13:12:25.579504  3624 net.cpp:406] icp2_reduction1 <- icp2_in_icp2_in_0_split_0
I0809 13:12:25.579519  3624 net.cpp:380] icp2_reduction1 -> icp2_reduction1
I0809 13:12:25.580039  3624 net.cpp:122] Setting up icp2_reduction1
I0809 13:12:25.580895  3624 net.cpp:129] Top shape: 1 128 120 160 (2457600)
I0809 13:12:25.580935  3624 net.cpp:137] Memory required for data: 543130212
I0809 13:12:25.580951  3624 layer_factory.hpp:77] Creating layer relu_icp2_reduction1
I0809 13:12:25.580981  3624 net.cpp:84] Creating Layer relu_icp2_reduction1
I0809 13:12:25.580991  3624 net.cpp:406] relu_icp2_reduction1 <- icp2_reduction1
I0809 13:12:25.581017  3624 net.cpp:367] relu_icp2_reduction1 -> icp2_reduction1 (in-place)
I0809 13:12:25.581033  3624 net.cpp:122] Setting up relu_icp2_reduction1
I0809 13:12:25.581058  3624 net.cpp:129] Top shape: 1 128 120 160 (2457600)
I0809 13:12:25.581066  3624 net.cpp:137] Memory required for data: 552960612
I0809 13:12:25.581074  3624 layer_factory.hpp:77] Creating layer icp2_reduction2
I0809 13:12:25.581105  3624 net.cpp:84] Creating Layer icp2_reduction2
I0809 13:12:25.581115  3624 net.cpp:406] icp2_reduction2 <- icp2_in_icp2_in_0_split_1
I0809 13:12:25.581142  3624 net.cpp:380] icp2_reduction2 -> icp2_reduction2
I0809 13:12:25.581928  3624 net.cpp:122] Setting up icp2_reduction2
I0809 13:12:25.581996  3624 net.cpp:129] Top shape: 1 32 120 160 (614400)
I0809 13:12:25.582005  3624 net.cpp:137] Memory required for data: 555418212
I0809 13:12:25.582051  3624 layer_factory.hpp:77] Creating layer relu_icp2_reduction2
I0809 13:12:25.582070  3624 net.cpp:84] Creating Layer relu_icp2_reduction2
I0809 13:12:25.582096  3624 net.cpp:406] relu_icp2_reduction2 <- icp2_reduction2
I0809 13:12:25.582113  3624 net.cpp:367] relu_icp2_reduction2 -> icp2_reduction2 (in-place)
I0809 13:12:25.582134  3624 net.cpp:122] Setting up relu_icp2_reduction2
I0809 13:12:25.582147  3624 net.cpp:129] Top shape: 1 32 120 160 (614400)
I0809 13:12:25.582159  3624 net.cpp:137] Memory required for data: 557875812
I0809 13:12:25.582170  3624 layer_factory.hpp:77] Creating layer icp2_pool
I0809 13:12:25.582186  3624 net.cpp:84] Creating Layer icp2_pool
I0809 13:12:25.582211  3624 net.cpp:406] icp2_pool <- icp2_in_icp2_in_0_split_2
I0809 13:12:25.582226  3624 net.cpp:380] icp2_pool -> icp2_pool
I0809 13:12:25.582250  3624 net.cpp:122] Setting up icp2_pool
I0809 13:12:25.582263  3624 net.cpp:129] Top shape: 1 256 120 160 (4915200)
I0809 13:12:25.582276  3624 net.cpp:137] Memory required for data: 577536612
I0809 13:12:25.582288  3624 layer_factory.hpp:77] Creating layer icp2_out0
I0809 13:12:25.582311  3624 net.cpp:84] Creating Layer icp2_out0
I0809 13:12:25.582336  3624 net.cpp:406] icp2_out0 <- icp2_in_icp2_in_0_split_3
I0809 13:12:25.582351  3624 net.cpp:380] icp2_out0 -> icp2_out0
I0809 13:12:25.583084  3624 net.cpp:122] Setting up icp2_out0
I0809 13:12:25.583168  3624 net.cpp:129] Top shape: 1 128 120 160 (2457600)
I0809 13:12:25.583176  3624 net.cpp:137] Memory required for data: 587367012
I0809 13:12:25.583194  3624 layer_factory.hpp:77] Creating layer relu_icp2_out0
I0809 13:12:25.583215  3624 net.cpp:84] Creating Layer relu_icp2_out0
I0809 13:12:25.583228  3624 net.cpp:406] relu_icp2_out0 <- icp2_out0
I0809 13:12:25.583245  3624 net.cpp:367] relu_icp2_out0 -> icp2_out0 (in-place)
I0809 13:12:25.583281  3624 net.cpp:122] Setting up relu_icp2_out0
I0809 13:12:25.583292  3624 net.cpp:129] Top shape: 1 128 120 160 (2457600)
I0809 13:12:25.583298  3624 net.cpp:137] Memory required for data: 597197412
I0809 13:12:25.583309  3624 layer_factory.hpp:77] Creating layer icp2_out1
I0809 13:12:25.583328  3624 net.cpp:84] Creating Layer icp2_out1
I0809 13:12:25.583353  3624 net.cpp:406] icp2_out1 <- icp2_reduction1
I0809 13:12:25.583387  3624 net.cpp:380] icp2_out1 -> icp2_out1
I0809 13:12:25.587761  3624 net.cpp:122] Setting up icp2_out1
I0809 13:12:25.587841  3624 net.cpp:129] Top shape: 1 192 120 160 (3686400)
I0809 13:12:25.587854  3624 net.cpp:137] Memory required for data: 611943012
I0809 13:12:25.587877  3624 layer_factory.hpp:77] Creating layer relu_icp2_out1
I0809 13:12:25.587904  3624 net.cpp:84] Creating Layer relu_icp2_out1
I0809 13:12:25.587921  3624 net.cpp:406] relu_icp2_out1 <- icp2_out1
I0809 13:12:25.587941  3624 net.cpp:367] relu_icp2_out1 -> icp2_out1 (in-place)
I0809 13:12:25.587965  3624 net.cpp:122] Setting up relu_icp2_out1
I0809 13:12:25.587996  3624 net.cpp:129] Top shape: 1 192 120 160 (3686400)
I0809 13:12:25.588011  3624 net.cpp:137] Memory required for data: 626688612
I0809 13:12:25.588021  3624 layer_factory.hpp:77] Creating layer icp2_out2
I0809 13:12:25.588045  3624 net.cpp:84] Creating Layer icp2_out2
I0809 13:12:25.588057  3624 net.cpp:406] icp2_out2 <- icp2_reduction2
I0809 13:12:25.588074  3624 net.cpp:380] icp2_out2 -> icp2_out2
I0809 13:12:25.590291  3624 net.cpp:122] Setting up icp2_out2
I0809 13:12:25.590667  3624 net.cpp:129] Top shape: 1 96 120 160 (1843200)
I0809 13:12:25.590679  3624 net.cpp:137] Memory required for data: 634061412
I0809 13:12:25.590695  3624 layer_factory.hpp:77] Creating layer relu_icp2_out2
I0809 13:12:25.590710  3624 net.cpp:84] Creating Layer relu_icp2_out2
I0809 13:12:25.590721  3624 net.cpp:406] relu_icp2_out2 <- icp2_out2
I0809 13:12:25.590750  3624 net.cpp:367] relu_icp2_out2 -> icp2_out2 (in-place)
I0809 13:12:25.590770  3624 net.cpp:122] Setting up relu_icp2_out2
I0809 13:12:25.590814  3624 net.cpp:129] Top shape: 1 96 120 160 (1843200)
I0809 13:12:25.590822  3624 net.cpp:137] Memory required for data: 641434212
I0809 13:12:25.590872  3624 layer_factory.hpp:77] Creating layer icp2_out3
I0809 13:12:25.590898  3624 net.cpp:84] Creating Layer icp2_out3
I0809 13:12:25.590936  3624 net.cpp:406] icp2_out3 <- icp2_pool
I0809 13:12:25.590968  3624 net.cpp:380] icp2_out3 -> icp2_out3
I0809 13:12:25.591259  3624 net.cpp:122] Setting up icp2_out3
I0809 13:12:25.591295  3624 net.cpp:129] Top shape: 1 64 120 160 (1228800)
I0809 13:12:25.591302  3624 net.cpp:137] Memory required for data: 646349412
I0809 13:12:25.591315  3624 layer_factory.hpp:77] Creating layer relu_icp2_out3
I0809 13:12:25.591325  3624 net.cpp:84] Creating Layer relu_icp2_out3
I0809 13:12:25.591333  3624 net.cpp:406] relu_icp2_out3 <- icp2_out3
I0809 13:12:25.591346  3624 net.cpp:367] relu_icp2_out3 -> icp2_out3 (in-place)
I0809 13:12:25.591358  3624 net.cpp:122] Setting up relu_icp2_out3
I0809 13:12:25.591368  3624 net.cpp:129] Top shape: 1 64 120 160 (1228800)
I0809 13:12:25.591393  3624 net.cpp:137] Memory required for data: 651264612
I0809 13:12:25.591401  3624 layer_factory.hpp:77] Creating layer icp2_out
I0809 13:12:25.591425  3624 net.cpp:84] Creating Layer icp2_out
I0809 13:12:25.591436  3624 net.cpp:406] icp2_out <- icp2_out0
I0809 13:12:25.591445  3624 net.cpp:406] icp2_out <- icp2_out1
I0809 13:12:25.591454  3624 net.cpp:406] icp2_out <- icp2_out2
I0809 13:12:25.591466  3624 net.cpp:406] icp2_out <- icp2_out3
I0809 13:12:25.591480  3624 net.cpp:380] icp2_out -> icp2_out
I0809 13:12:25.591493  3624 net.cpp:122] Setting up icp2_out
I0809 13:12:25.591506  3624 net.cpp:129] Top shape: 1 480 120 160 (9216000)
I0809 13:12:25.591513  3624 net.cpp:137] Memory required for data: 688128612
I0809 13:12:25.591521  3624 layer_factory.hpp:77] Creating layer icp3_in
I0809 13:12:25.591538  3624 net.cpp:84] Creating Layer icp3_in
I0809 13:12:25.591549  3624 net.cpp:406] icp3_in <- icp2_out
I0809 13:12:25.591560  3624 net.cpp:380] icp3_in -> icp3_in
I0809 13:12:25.591583  3624 net.cpp:122] Setting up icp3_in
I0809 13:12:25.591610  3624 net.cpp:129] Top shape: 1 480 60 80 (2304000)
I0809 13:12:25.591619  3624 net.cpp:137] Memory required for data: 697344612
I0809 13:12:25.591626  3624 layer_factory.hpp:77] Creating layer icp3_in_icp3_in_0_split
I0809 13:12:25.591637  3624 net.cpp:84] Creating Layer icp3_in_icp3_in_0_split
I0809 13:12:25.591646  3624 net.cpp:406] icp3_in_icp3_in_0_split <- icp3_in
I0809 13:12:25.591665  3624 net.cpp:380] icp3_in_icp3_in_0_split -> icp3_in_icp3_in_0_split_0
I0809 13:12:25.591681  3624 net.cpp:380] icp3_in_icp3_in_0_split -> icp3_in_icp3_in_0_split_1
I0809 13:12:25.591693  3624 net.cpp:380] icp3_in_icp3_in_0_split -> icp3_in_icp3_in_0_split_2
I0809 13:12:25.591709  3624 net.cpp:380] icp3_in_icp3_in_0_split -> icp3_in_icp3_in_0_split_3
I0809 13:12:25.591725  3624 net.cpp:122] Setting up icp3_in_icp3_in_0_split
I0809 13:12:25.591738  3624 net.cpp:129] Top shape: 1 480 60 80 (2304000)
I0809 13:12:25.591747  3624 net.cpp:129] Top shape: 1 480 60 80 (2304000)
I0809 13:12:25.591756  3624 net.cpp:129] Top shape: 1 480 60 80 (2304000)
I0809 13:12:25.591768  3624 net.cpp:129] Top shape: 1 480 60 80 (2304000)
I0809 13:12:25.591775  3624 net.cpp:137] Memory required for data: 734208612
I0809 13:12:25.591783  3624 layer_factory.hpp:77] Creating layer icp3_reduction1
I0809 13:12:25.591800  3624 net.cpp:84] Creating Layer icp3_reduction1
I0809 13:12:25.591809  3624 net.cpp:406] icp3_reduction1 <- icp3_in_icp3_in_0_split_0
I0809 13:12:25.591823  3624 net.cpp:380] icp3_reduction1 -> icp3_reduction1
I0809 13:12:25.592313  3624 net.cpp:122] Setting up icp3_reduction1
I0809 13:12:25.592347  3624 net.cpp:129] Top shape: 1 96 60 80 (460800)
I0809 13:12:25.592356  3624 net.cpp:137] Memory required for data: 736051812
I0809 13:12:25.592370  3624 layer_factory.hpp:77] Creating layer relu_icp3_reduction1
I0809 13:12:25.592384  3624 net.cpp:84] Creating Layer relu_icp3_reduction1
I0809 13:12:25.592393  3624 net.cpp:406] relu_icp3_reduction1 <- icp3_reduction1
I0809 13:12:25.592406  3624 net.cpp:367] relu_icp3_reduction1 -> icp3_reduction1 (in-place)
I0809 13:12:25.592420  3624 net.cpp:122] Setting up relu_icp3_reduction1
I0809 13:12:25.592430  3624 net.cpp:129] Top shape: 1 96 60 80 (460800)
I0809 13:12:25.592437  3624 net.cpp:137] Memory required for data: 737895012
I0809 13:12:25.592444  3624 layer_factory.hpp:77] Creating layer icp3_reduction2
I0809 13:12:25.592463  3624 net.cpp:84] Creating Layer icp3_reduction2
I0809 13:12:25.592471  3624 net.cpp:406] icp3_reduction2 <- icp3_in_icp3_in_0_split_1
I0809 13:12:25.592484  3624 net.cpp:380] icp3_reduction2 -> icp3_reduction2
I0809 13:12:25.592597  3624 net.cpp:122] Setting up icp3_reduction2
I0809 13:12:25.592633  3624 net.cpp:129] Top shape: 1 16 60 80 (76800)
I0809 13:12:25.592643  3624 net.cpp:137] Memory required for data: 738202212
I0809 13:12:25.592669  3624 layer_factory.hpp:77] Creating layer relu_icp3_reduction2
I0809 13:12:25.592685  3624 net.cpp:84] Creating Layer relu_icp3_reduction2
I0809 13:12:25.592694  3624 net.cpp:406] relu_icp3_reduction2 <- icp3_reduction2
I0809 13:12:25.592707  3624 net.cpp:367] relu_icp3_reduction2 -> icp3_reduction2 (in-place)
I0809 13:12:25.592720  3624 net.cpp:122] Setting up relu_icp3_reduction2
I0809 13:12:25.592733  3624 net.cpp:129] Top shape: 1 16 60 80 (76800)
I0809 13:12:25.592743  3624 net.cpp:137] Memory required for data: 738509412
I0809 13:12:25.592751  3624 layer_factory.hpp:77] Creating layer icp3_pool
I0809 13:12:25.592764  3624 net.cpp:84] Creating Layer icp3_pool
I0809 13:12:25.592775  3624 net.cpp:406] icp3_pool <- icp3_in_icp3_in_0_split_2
I0809 13:12:25.592787  3624 net.cpp:380] icp3_pool -> icp3_pool
I0809 13:12:25.592808  3624 net.cpp:122] Setting up icp3_pool
I0809 13:12:25.592838  3624 net.cpp:129] Top shape: 1 480 60 80 (2304000)
I0809 13:12:25.592845  3624 net.cpp:137] Memory required for data: 747725412
I0809 13:12:25.592854  3624 layer_factory.hpp:77] Creating layer icp3_out0
I0809 13:12:25.592867  3624 net.cpp:84] Creating Layer icp3_out0
I0809 13:12:25.592875  3624 net.cpp:406] icp3_out0 <- icp3_in_icp3_in_0_split_3
I0809 13:12:25.592891  3624 net.cpp:380] icp3_out0 -> icp3_out0
I0809 13:12:25.593799  3624 net.cpp:122] Setting up icp3_out0
I0809 13:12:25.593951  3624 net.cpp:129] Top shape: 1 192 60 80 (921600)
I0809 13:12:25.593963  3624 net.cpp:137] Memory required for data: 751411812
I0809 13:12:25.594000  3624 layer_factory.hpp:77] Creating layer relu_icp3_out0
I0809 13:12:25.594024  3624 net.cpp:84] Creating Layer relu_icp3_out0
I0809 13:12:25.594069  3624 net.cpp:406] relu_icp3_out0 <- icp3_out0
I0809 13:12:25.594092  3624 net.cpp:367] relu_icp3_out0 -> icp3_out0 (in-place)
I0809 13:12:25.594110  3624 net.cpp:122] Setting up relu_icp3_out0
I0809 13:12:25.594120  3624 net.cpp:129] Top shape: 1 192 60 80 (921600)
I0809 13:12:25.594127  3624 net.cpp:137] Memory required for data: 755098212
I0809 13:12:25.594135  3624 layer_factory.hpp:77] Creating layer icp3_out1
I0809 13:12:25.594161  3624 net.cpp:84] Creating Layer icp3_out1
I0809 13:12:25.594174  3624 net.cpp:406] icp3_out1 <- icp3_reduction1
I0809 13:12:25.594190  3624 net.cpp:380] icp3_out1 -> icp3_out1
I0809 13:12:25.597784  3624 net.cpp:122] Setting up icp3_out1
I0809 13:12:25.597893  3624 net.cpp:129] Top shape: 1 208 60 80 (998400)
I0809 13:12:25.597940  3624 net.cpp:137] Memory required for data: 759091812
I0809 13:12:25.597959  3624 layer_factory.hpp:77] Creating layer relu_icp3_out1
I0809 13:12:25.597993  3624 net.cpp:84] Creating Layer relu_icp3_out1
I0809 13:12:25.598033  3624 net.cpp:406] relu_icp3_out1 <- icp3_out1
I0809 13:12:25.598047  3624 net.cpp:367] relu_icp3_out1 -> icp3_out1 (in-place)
I0809 13:12:25.598078  3624 net.cpp:122] Setting up relu_icp3_out1
I0809 13:12:25.598088  3624 net.cpp:129] Top shape: 1 208 60 80 (998400)
I0809 13:12:25.598109  3624 net.cpp:137] Memory required for data: 763085412
I0809 13:12:25.598117  3624 layer_factory.hpp:77] Creating layer icp3_out2
I0809 13:12:25.598134  3624 net.cpp:84] Creating Layer icp3_out2
I0809 13:12:25.598158  3624 net.cpp:406] icp3_out2 <- icp3_reduction2
I0809 13:12:25.598173  3624 net.cpp:380] icp3_out2 -> icp3_out2
I0809 13:12:25.598564  3624 net.cpp:122] Setting up icp3_out2
I0809 13:12:25.598629  3624 net.cpp:129] Top shape: 1 48 60 80 (230400)
I0809 13:12:25.598639  3624 net.cpp:137] Memory required for data: 764007012
I0809 13:12:25.598654  3624 layer_factory.hpp:77] Creating layer relu_icp3_out2
I0809 13:12:25.598685  3624 net.cpp:84] Creating Layer relu_icp3_out2
I0809 13:12:25.598695  3624 net.cpp:406] relu_icp3_out2 <- icp3_out2
I0809 13:12:25.598721  3624 net.cpp:367] relu_icp3_out2 -> icp3_out2 (in-place)
I0809 13:12:25.598733  3624 net.cpp:122] Setting up relu_icp3_out2
I0809 13:12:25.598758  3624 net.cpp:129] Top shape: 1 48 60 80 (230400)
I0809 13:12:25.598767  3624 net.cpp:137] Memory required for data: 764928612
I0809 13:12:25.598774  3624 layer_factory.hpp:77] Creating layer icp3_out3
I0809 13:12:25.598803  3624 net.cpp:84] Creating Layer icp3_out3
I0809 13:12:25.598815  3624 net.cpp:406] icp3_out3 <- icp3_pool
I0809 13:12:25.598845  3624 net.cpp:380] icp3_out3 -> icp3_out3
I0809 13:12:25.599194  3624 net.cpp:122] Setting up icp3_out3
I0809 13:12:25.599232  3624 net.cpp:129] Top shape: 1 64 60 80 (307200)
I0809 13:12:25.599241  3624 net.cpp:137] Memory required for data: 766157412
I0809 13:12:25.599253  3624 layer_factory.hpp:77] Creating layer relu_icp3_out3
I0809 13:12:25.599267  3624 net.cpp:84] Creating Layer relu_icp3_out3
I0809 13:12:25.599279  3624 net.cpp:406] relu_icp3_out3 <- icp3_out3
I0809 13:12:25.599292  3624 net.cpp:367] relu_icp3_out3 -> icp3_out3 (in-place)
I0809 13:12:25.599303  3624 net.cpp:122] Setting up relu_icp3_out3
I0809 13:12:25.599313  3624 net.cpp:129] Top shape: 1 64 60 80 (307200)
I0809 13:12:25.599323  3624 net.cpp:137] Memory required for data: 767386212
I0809 13:12:25.599330  3624 layer_factory.hpp:77] Creating layer icp4_in
I0809 13:12:25.599340  3624 net.cpp:84] Creating Layer icp4_in
I0809 13:12:25.599351  3624 net.cpp:406] icp4_in <- icp3_out0
I0809 13:12:25.599362  3624 net.cpp:406] icp4_in <- icp3_out1
I0809 13:12:25.599371  3624 net.cpp:406] icp4_in <- icp3_out2
I0809 13:12:25.599380  3624 net.cpp:406] icp4_in <- icp3_out3
I0809 13:12:25.599395  3624 net.cpp:380] icp4_in -> icp4_in
I0809 13:12:25.599408  3624 net.cpp:122] Setting up icp4_in
I0809 13:12:25.599421  3624 net.cpp:129] Top shape: 1 512 60 80 (2457600)
I0809 13:12:25.599428  3624 net.cpp:137] Memory required for data: 777216612
I0809 13:12:25.599436  3624 layer_factory.hpp:77] Creating layer lstm1
I0809 13:12:25.599450  3624 net.cpp:84] Creating Layer lstm1
I0809 13:12:25.599458  3624 net.cpp:406] lstm1 <- icp4_in
I0809 13:12:25.599467  3624 net.cpp:406] lstm1 <- clip_input_0_split_0
I0809 13:12:25.599483  3624 net.cpp:380] lstm1 -> lstm1
I0809 13:12:25.599498  3624 recurrent_layer.cpp:20] Initializing recurrent layer: assuming input batch contains 1 timesteps of 512 independent streams.
F0809 13:12:25.599516  3624 recurrent_layer.cpp:26] Check failed: N_ == bottom[1]->shape(1) (512 vs. 51) 
*** Check failure stack trace: ***
Aborted (core dumped)

Reply all

Reply to author

Forward

0 new messages