faceboxes_deploy.prototxt

name: "FaceBoxes"
input: "data"
input_shape {
  dim: 1
  dim: 3
  dim: 1024
  dim: 1024
}

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: 24
    kernel_size: 7
    stride: 4
    pad: 3
    bias_term: true
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "conv1bn"
  type: "BatchNorm"
  bottom: "conv1"
  top: "conv1"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
}
layer {
  name: "conv1neg"
  type: "Power"
  bottom: "conv1"
  top: "conv1neg"
  power_param {
    power: 1
    scale: -1.0
    shift: 0
  }
}
layer {
  name: "conv1concat"
  type: "Concat"
  bottom: "conv1"
  bottom: "conv1neg"
  top: "conv1_1"
}
layer {
  name: "conv1scale"
  type: "Scale"
  bottom: "conv1_1"
  top: "conv1_1"
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  param {
    lr_mult: 0.2
    decay_mult: 0
  }
  scale_param {
    bias_term: true
  }
}
layer {
  name: "conv1relu"
  type: "ReLU"
  bottom: "conv1_1"
  top: "conv1_1"
}
layer {
  name: "pool1"
  type: "Pooling"
  bottom: "conv1_1"
  top: "pool1"
  pooling_param {
    pool: MAX
    kernel_size: 3
    stride: 2
    pad: 0
  }
}

###########################conv2##############################

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: 64
    kernel_size: 5
    stride: 2
    pad: 3
    bias_term: true
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "conv2bn"
  type: "BatchNorm"
  bottom: "conv2"
  top: "conv2"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
}
layer {
  name: "conv2neg"
  type: "Power"
  bottom: "conv2"
  top: "conv2neg"
  power_param {
    power: 1
    scale: -1.0
    shift: 0
  }
}
layer {
  name: "conv2concat"
  type: "Concat"
  bottom: "conv2"
  bottom: "conv2neg"
  top: "conv2_1"
}
layer {
  name: "conv2scale"
  type: "Scale"
  bottom: "conv2_1"
  top: "conv2_1"
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  param {
    lr_mult: 0.2
    decay_mult: 0
  }
  scale_param {
    bias_term: true
  }
}
layer {
  name: "conv2relu"
  type: "ReLU"
  bottom: "conv2_1"
  top: "conv2_1"
}
layer {
  name: "pool2"
  type: "Pooling"
  bottom: "conv2_1"
  top: "pool2"
  pooling_param {
    pool: MAX
    kernel_size: 3
    stride: 2
    pad: 0
  }
}


##########################inception--1##############################

layer {
  name: "conv2_1/incep/0/conv"
  type: "Convolution"
  bottom: "pool2"
  top: "conv2_1/incep/0"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 32
    kernel_size: 1
    stride: 1
    pad: 0
    bias_term: true
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "conv2_1/incep/0/bn"
  type: "BatchNorm"
  bottom: "conv2_1/incep/0"
  top: "conv2_1/incep/0"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
}
layer {
  name: "conv2_1/incep/0/bn_scale"
  type: "Scale"
  bottom: "conv2_1/incep/0"
  top: "conv2_1/incep/0"
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  scale_param {
    bias_term: true
  }
}
layer {
  name: "conv2_1/incep/0/relu"
  type: "ReLU"
  bottom: "conv2_1/incep/0"
  top: "conv2_1/incep/0"
}

layer {
  name: "conv2_1/incep/1_reduce/conv"
  type: "Convolution"
  bottom: "pool2"
  top: "conv2_1/incep/1_reduce"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 24
    kernel_size: 1
    stride: 1
    bias_term: true
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "conv2_1/incep/1_reduce/bn"
  type: "BatchNorm"
  bottom: "conv2_1/incep/1_reduce"
  top: "conv2_1/incep/1_reduce"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
}
layer {
  name: "conv2_1/incep/1_reduce/bn_scale"
  type: "Scale"
  bottom: "conv2_1/incep/1_reduce"
  top: "conv2_1/incep/1_reduce"
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  scale_param {
    bias_term: true
  }
}
layer {
  name: "conv2_1/incep/1_reduce/relu"
  type: "ReLU"
  bottom: "conv2_1/incep/1_reduce"
  top: "conv2_1/incep/1_reduce"
}
layer {
  name: "conv2_1/incep/1_0/conv"
  type: "Convolution"
  bottom: "conv2_1/incep/1_reduce"
  top: "conv2_1/incep/1_0"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 32
    kernel_size: 3
    stride: 1
    pad: 1
    bias_term: true
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "conv2_1/incep/1_0/bn"
  type: "BatchNorm"
  bottom: "conv2_1/incep/1_0"
  top: "conv2_1/incep/1_0"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }

}
layer {
  name: "conv2_1/incep/1_0/bn_scale"
  type: "Scale"
  bottom: "conv2_1/incep/1_0"
  top: "conv2_1/incep/1_0"
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  scale_param {
    bias_term: true
  }
}
layer {
  name: "conv2_1/incep/1_0/relu"
  type: "ReLU"
  bottom: "conv2_1/incep/1_0"
  top: "conv2_1/incep/1_0"
}

###############################

layer {
  name: "conv2_1/incep/2_reduce/conv"
  type: "Convolution"
  bottom: "pool2"
  top: "conv2_1/incep/2_reduce"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 24
    kernel_size: 1
    stride: 1
    pad: 0
    bias_term: true
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "conv2_1/incep/2_reduce/bn"
  type: "BatchNorm"
  bottom: "conv2_1/incep/2_reduce"
  top: "conv2_1/incep/2_reduce"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
}
layer {
  name: "conv2_1/incep/2_reduce/bn_scale"
  type: "Scale"
  bottom: "conv2_1/incep/2_reduce"
  top: "conv2_1/incep/2_reduce"
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  scale_param {
    bias_term: true
  }
}
layer {
  name: "conv2_1/incep/2_reduce/relu"
  type: "ReLU"
  bottom: "conv2_1/incep/2_reduce"
  top: "conv2_1/incep/2_reduce"
}
layer {
  name: "conv2_1/incep/2_0/conv"
  type: "Convolution"
  bottom: "conv2_1/incep/2_reduce"
  top: "conv2_1/incep/2_0"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 32
    kernel_size: 3
    stride: 1
    pad: 1
    bias_term: true
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "conv2_1/incep/2_0/bn"
  type: "BatchNorm"
  bottom: "conv2_1/incep/2_0"
  top: "conv2_1/incep/2_0"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
}
layer {
  name: "conv2_1/incep/2_0/bn_scale"
  type: "Scale"
  bottom: "conv2_1/incep/2_0"
  top: "conv2_1/incep/2_0"
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  scale_param {
    bias_term: true
  }
}
layer {
  name: "conv2_1/incep/2_0/relu"
  type: "ReLU"
  bottom: "conv2_1/incep/2_0"
  top: "conv2_1/incep/2_0"
}
layer {
  name: "conv2_1/incep/2_1/conv"
  type: "Convolution"
  bottom: "conv2_1/incep/2_0"
  top: "conv2_1/incep/2_1"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 32
    kernel_size: 3
    stride: 1
    pad: 1
    bias_term: true
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "conv2_1/incep/2_1/bn"
  type: "BatchNorm"
  bottom: "conv2_1/incep/2_1"
  top: "conv2_1/incep/2_1"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
}
layer {
  name: "conv2_1/incep/2_1/bn_scale"
  type: "Scale"
  bottom: "conv2_1/incep/2_1"
  top: "conv2_1/incep/2_1"
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  scale_param {
    bias_term: true
  }
}
layer {
  name: "conv2_1/incep/2_1/relu"
  type: "ReLU"
  bottom: "conv2_1/incep/2_1"
  top: "conv2_1/incep/2_1"
}

##########################################

layer {
  name: "conv2_1/incep/pool"
  type: "Pooling"
  bottom: "pool2"
  top: "conv2_1/incep/pool"
  pooling_param {
    pool: MAX
    kernel_size: 3
    stride: 1
    pad: 1
  }
}
layer {
  name: "conv2_1/incep/poolproj/conv"
  type: "Convolution"
  bottom: "conv2_1/incep/pool"
  top: "conv2_1/incep/poolproj"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 32
    kernel_size: 1
    stride: 1
    pad: 0
    bias_term: true
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "conv2_1/incep/poolproj/bn"
  type: "BatchNorm"
  bottom: "conv2_1/incep/poolproj"
  top: "conv2_1/incep/poolproj"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
}
layer {
  name: "conv2_1/incep/poolproj/bn_scale"
  type: "Scale"
  bottom: "conv2_1/incep/poolproj"
  top: "conv2_1/incep/poolproj"
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  scale_param {
    bias_term: true
  }
}
layer {
  name: "conv2_1/incep/poolproj/relu"
  type: "ReLU"
  bottom: "conv2_1/incep/poolproj"
  top: "conv2_1/incep/poolproj"
}
layer {
  name: "conv2_1/incep"
  type: "Concat"
  bottom: "conv2_1/incep/0"
  bottom: "conv2_1/incep/1_0"
  bottom: "conv2_1/incep/2_1"
  bottom: "conv2_1/incep/poolproj"
  top: "conv2_1/incep"
}


##########################inception--2##############################

layer {
  name: "conv2_2/incep/0/conv"
  type: "Convolution"
  bottom: "conv2_1/incep"
  top: "conv2_2/incep/0"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 32
    kernel_size: 1
    stride: 1
    pad: 0
    bias_term: true
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "conv2_2/incep/0/bn"
  type: "BatchNorm"
  bottom: "conv2_2/incep/0"
  top: "conv2_2/incep/0"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
}
layer {
  name: "conv2_2/incep/0/bn_scale"
  type: "Scale"
  bottom: "conv2_2/incep/0"
  top: "conv2_2/incep/0"
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  scale_param {
    bias_term: true
  }
}
layer {
  name: "conv2_2/incep/0/relu"
  type: "ReLU"
  bottom: "conv2_2/incep/0"
  top: "conv2_2/incep/0"
}

layer {
  name: "conv2_2/incep/1_reduce/conv"
  type: "Convolution"
  bottom: "conv2_1/incep"
  top: "conv2_2/incep/1_reduce"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 24
    kernel_size: 1
    stride: 1
    bias_term: true
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "conv2_2/incep/1_reduce/bn"
  type: "BatchNorm"
  bottom: "conv2_2/incep/1_reduce"
  top: "conv2_2/incep/1_reduce"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
}
layer {
  name: "conv2_2/incep/1_reduce/bn_scale"
  type: "Scale"
  bottom: "conv2_2/incep/1_reduce"
  top: "conv2_2/incep/1_reduce"
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  scale_param {
    bias_term: true
  }
}
layer {
  name: "conv2_2/incep/1_reduce/relu"
  type: "ReLU"
  bottom: "conv2_2/incep/1_reduce"
  top: "conv2_2/incep/1_reduce"
}
layer {
  name: "conv2_2/incep/1_0/conv"
  type: "Convolution"
  bottom: "conv2_2/incep/1_reduce"
  top: "conv2_2/incep/1_0"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 32
    kernel_size: 3
    stride: 1
    pad: 1
    bias_term: true
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "conv2_2/incep/1_0/bn"
  type: "BatchNorm"
  bottom: "conv2_2/incep/1_0"
  top: "conv2_2/incep/1_0"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
}
layer {
  name: "conv2_2/incep/1_0/bn_scale"
  type: "Scale"
  bottom: "conv2_2/incep/1_0"
  top: "conv2_2/incep/1_0"
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  scale_param {
    bias_term: true
  }
}
layer {
  name: "conv2_2/incep/1_0/relu"
  type: "ReLU"
  bottom: "conv2_2/incep/1_0"
  top: "conv2_2/incep/1_0"
}

###############################

layer {
  name: "conv2_2/incep/2_reduce/conv"
  type: "Convolution"
  bottom: "conv2_1/incep"
  top: "conv2_2/incep/2_reduce"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 24
    kernel_size: 1
    stride: 1
    pad: 0
    bias_term: true
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "conv2_2/incep/2_reduce/bn"
  type: "BatchNorm"
  bottom: "conv2_2/incep/2_reduce"
  top: "conv2_2/incep/2_reduce"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
}
layer {
  name: "conv2_2/incep/2_reduce/bn_scale"
  type: "Scale"
  bottom: "conv2_2/incep/2_reduce"
  top: "conv2_2/incep/2_reduce"
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  scale_param {
    bias_term: true
  }
}
layer {
  name: "conv2_2/incep/2_reduce/relu"
  type: "ReLU"
  bottom: "conv2_2/incep/2_reduce"
  top: "conv2_2/incep/2_reduce"
}
layer {
  name: "conv2_2/incep/2_0/conv"
  type: "Convolution"
  bottom: "conv2_2/incep/2_reduce"
  top: "conv2_2/incep/2_0"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 32
    kernel_size: 3
    stride: 1
    pad: 1
    bias_term: true
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "conv2_2/incep/2_0/bn"
  type: "BatchNorm"
  bottom: "conv2_2/incep/2_0"
  top: "conv2_2/incep/2_0"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
}
layer {
  name: "conv2_2/incep/2_0/bn_scale"
  type: "Scale"
  bottom: "conv2_2/incep/2_0"
  top: "conv2_2/incep/2_0"
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  scale_param {
    bias_term: true
  }
}
layer {
  name: "conv2_2/incep/2_0/relu"
  type: "ReLU"
  bottom: "conv2_2/incep/2_0"
  top: "conv2_2/incep/2_0"
}
layer {
  name: "conv2_2/incep/2_1/conv"
  type: "Convolution"
  bottom: "conv2_2/incep/2_0"
  top: "conv2_2/incep/2_1"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 32
    kernel_size: 3
    stride: 1
    pad: 1
    bias_term: true
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "conv2_2/incep/2_1/bn"
  type: "BatchNorm"
  bottom: "conv2_2/incep/2_1"
  top: "conv2_2/incep/2_1"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
}
layer {
  name: "conv2_2/incep/2_1/bn_scale"
  type: "Scale"
  bottom: "conv2_2/incep/2_1"
  top: "conv2_2/incep/2_1"
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  scale_param {
    bias_term: true
  }
}
layer {
  name: "conv2_2/incep/2_1/relu"
  type: "ReLU"
  bottom: "conv2_2/incep/2_1"
  top: "conv2_2/incep/2_1"
}

##########################################

layer {
  name: "conv2_2/incep/pool"
  type: "Pooling"
  bottom: "conv2_1/incep"
  top: "conv2_2/incep/pool"
  pooling_param {
    pool: MAX
    kernel_size: 3
    stride: 1
    pad: 1
  }
}
layer {
  name: "conv2_2/incep/poolproj/conv"
  type: "Convolution"
  bottom: "conv2_2/incep/pool"
  top: "conv2_2/incep/poolproj"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 32
    kernel_size: 1
    stride: 1
    pad: 0
    bias_term: true
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "conv2_2/incep/poolproj/bn"
  type: "BatchNorm"
  bottom: "conv2_2/incep/poolproj"
  top: "conv2_2/incep/poolproj"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
}
layer {
  name: "conv2_2/incep/poolproj/bn_scale"
  type: "Scale"
  bottom: "conv2_2/incep/poolproj"
  top: "conv2_2/incep/poolproj"
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  scale_param {
    bias_term: true
  }
}
layer {
  name: "conv2_2/incep/poolproj/relu"
  type: "ReLU"
  bottom: "conv2_2/incep/poolproj"
  top: "conv2_2/incep/poolproj"
}
layer {
  name: "conv2_2/incep"
  type: "Concat"
  bottom: "conv2_2/incep/0"
  bottom: "conv2_2/incep/1_0"
  bottom: "conv2_2/incep/2_1"
  bottom: "conv2_2/incep/poolproj"
  top: "conv2_2/incep"
}

##########################inception--3##############################

layer {
  name: "conv2_3/incep/0/conv"
  type: "Convolution"
  bottom: "conv2_2/incep"
  top: "conv2_3/incep/0"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 32
    kernel_size: 1
    stride: 1
    pad: 0
    bias_term: true
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "conv2_3/incep/0/bn"
  type: "BatchNorm"
  bottom: "conv2_3/incep/0"
  top: "conv2_3/incep/0"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
}
layer {
  name: "conv2_3/incep/0/bn_scale"
  type: "Scale"
  bottom: "conv2_3/incep/0"
  top: "conv2_3/incep/0"
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  scale_param {
    bias_term: true
  }
}
layer {
  name: "conv2_3/incep/0/relu"
  type: "ReLU"
  bottom: "conv2_3/incep/0"
  top: "conv2_3/incep/0"
}
layer {
  name: "conv2_3/incep/1_reduce/conv"
  type: "Convolution"
  bottom: "conv2_2/incep"
  top: "conv2_3/incep/1_reduce"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 24
    kernel_size: 1
    stride: 1
    bias_term: true
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "conv2_3/incep/1_reduce/bn"
  type: "BatchNorm"
  bottom: "conv2_3/incep/1_reduce"
  top: "conv2_3/incep/1_reduce"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
}
layer {
  name: "conv2_3/incep/1_reduce/bn_scale"
  type: "Scale"
  bottom: "conv2_3/incep/1_reduce"
  top: "conv2_3/incep/1_reduce"
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  scale_param {
    bias_term: true
  }
}
layer {
  name: "conv2_3/incep/1_reduce/relu"
  type: "ReLU"
  bottom: "conv2_3/incep/1_reduce"
  top: "conv2_3/incep/1_reduce"
}
layer {
  name: "conv2_3/incep/1_0/conv"
  type: "Convolution"
  bottom: "conv2_3/incep/1_reduce"
  top: "conv2_3/incep/1_0"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 32
    kernel_size: 3
    stride: 1
    pad: 1
    bias_term: true
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "conv2_3/incep/1_0/bn"
  type: "BatchNorm"
  bottom: "conv2_3/incep/1_0"
  top: "conv2_3/incep/1_0"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
}
layer {
  name: "conv2_3/incep/1_0/bn_scale"
  type: "Scale"
  bottom: "conv2_3/incep/1_0"
  top: "conv2_3/incep/1_0"
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  scale_param {
    bias_term: true
  }
}
layer {
  name: "conv2_3/incep/1_0/relu"
  type: "ReLU"
  bottom: "conv2_3/incep/1_0"
  top: "conv2_3/incep/1_0"
}

###############################

layer {
  name: "conv2_3/incep/2_reduce/conv"
  type: "Convolution"
  bottom: "conv2_2/incep"
  top: "conv2_3/incep/2_reduce"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 24
    kernel_size: 1
    stride: 1
    pad: 0
    bias_term: true
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "conv2_3/incep/2_reduce/bn"
  type: "BatchNorm"
  bottom: "conv2_3/incep/2_reduce"
  top: "conv2_3/incep/2_reduce"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
}
layer {
  name: "conv2_3/incep/2_reduce/bn_scale"
  type: "Scale"
  bottom: "conv2_3/incep/2_reduce"
  top: "conv2_3/incep/2_reduce"
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  scale_param {
    bias_term: true
  }
}
layer {
  name: "conv2_3/incep/2_reduce/relu"
  type: "ReLU"
  bottom: "conv2_3/incep/2_reduce"
  top: "conv2_3/incep/2_reduce"
}
layer {
  name: "conv2_3/incep/2_0/conv"
  type: "Convolution"
  bottom: "conv2_3/incep/2_reduce"
  top: "conv2_3/incep/2_0"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 32
    kernel_size: 3
    stride: 1
    pad: 1
    bias_term: true
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "conv2_3/incep/2_0/bn"
  type: "BatchNorm"
  bottom: "conv2_3/incep/2_0"
  top: "conv2_3/incep/2_0"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }

}
layer {
  name: "conv2_3/incep/2_0/bn_scale"
  type: "Scale"
  bottom: "conv2_3/incep/2_0"
  top: "conv2_3/incep/2_0"
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  scale_param {
    bias_term: true
  }
}
layer {
  name: "conv2_3/incep/2_0/relu"
  type: "ReLU"
  bottom: "conv2_3/incep/2_0"
  top: "conv2_3/incep/2_0"
}
layer {
  name: "conv2_3/incep/2_1/conv"
  type: "Convolution"
  bottom: "conv2_3/incep/2_0"
  top: "conv2_3/incep/2_1"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 32
    kernel_size: 3
    stride: 1
    pad: 1
    bias_term: true
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "conv2_3/incep/2_1/bn"
  type: "BatchNorm"
  bottom: "conv2_3/incep/2_1"
  top: "conv2_3/incep/2_1"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
}
layer {
  name: "conv2_3/incep/2_1/bn_scale"
  type: "Scale"
  bottom: "conv2_3/incep/2_1"
  top: "conv2_3/incep/2_1"
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  scale_param {
    bias_term: true
  }
}
layer {
  name: "conv2_3/incep/2_1/relu"
  type: "ReLU"
  bottom: "conv2_3/incep/2_1"
  top: "conv2_3/incep/2_1"
}

##########################################

layer {
  name: "conv2_3/incep/pool"
  type: "Pooling"
  bottom: "conv2_2/incep"
  top: "conv2_3/incep/pool"
  pooling_param {
    pool: MAX
    kernel_size: 3
    stride: 1
    pad: 1
  }
}
layer {
  name: "conv2_3/incep/poolproj/conv"
  type: "Convolution"
  bottom: "conv2_3/incep/pool"
  top: "conv2_3/incep/poolproj"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 32
    kernel_size: 1
    stride: 1
    pad: 0
    bias_term: true
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "conv2_3/incep/poolproj/bn"
  type: "BatchNorm"
  bottom: "conv2_3/incep/poolproj"
  top: "conv2_3/incep/poolproj"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
}
layer {
  name: "conv2_3/incep/poolproj/bn_scale"
  type: "Scale"
  bottom: "conv2_3/incep/poolproj"
  top: "conv2_3/incep/poolproj"
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  scale_param {
    bias_term: true
  }
}
layer {
  name: "conv2_3/incep/poolproj/relu"
  type: "ReLU"
  bottom: "conv2_3/incep/poolproj"
  top: "conv2_3/incep/poolproj"
}
layer {
  name: "conv2_3/incep"
  type: "Concat"
  bottom: "conv2_3/incep/0"
  bottom: "conv2_3/incep/1_0"
  bottom: "conv2_3/incep/2_1"
  bottom: "conv2_3/incep/poolproj"
  top: "conv2_3/incep"
}

layer {
  name: "inception3_priorbox"
  type: "PriorBox"
  bottom: "conv2_3/incep"
  bottom: "data"
  top: "inception3_priorbox"
  prior_box_param {
    fixed_size: 32
    fixed_size: 64
    fixed_size: 128
    density: 4
    density: 2
    density: 1
    step:32
    variance: 0.1
    variance: 0.1
    variance: 0.2
    variance: 0.2
    offset: 0.5
  }
}

layer {
  name: "inception3_loc"
  type: "Convolution"
  bottom: "conv2_3/incep"
  top: "inception3_loc"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 84
    pad: 1
    kernel_size: 3
    stride: 1
    bias_term: true
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "inception3_loc_perm"
  type: "Permute"
  bottom: "inception3_loc"
  top: "inception3_loc_perm"
  permute_param {
    order: 0
    order: 2
    order: 3
    order: 1
  }
}
layer {
  name: "inception3_loc_flat"
  type: "Flatten"
  bottom: "inception3_loc_perm"
  top: "inception3_loc_flat"
  flatten_param {
    axis: 1
  }
}

layer {
  name: "inception3_conf"
  type: "Convolution"
  bottom: "conv2_3/incep"
  top: "inception3_conf"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 42
    pad: 1
    kernel_size: 3
    stride: 1
    bias_term: true
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "inception3_conf_perm"
  type: "Permute"
  bottom: "inception3_conf"
  top: "inception3_conf_perm"
  permute_param {
    order: 0
    order: 2
    order: 3
    order: 1
  }
}
layer {
  name: "inception3_conf_flat"
  type: "Flatten"
  bottom: "inception3_conf_perm"
  top: "inception3_conf_flat"
  flatten_param {
    axis: 1
  }
}

##################################conv3#########################
layer {
  name: "conv3_1"
  type: "Convolution"
  bottom: "conv2_3/incep"
  top: "conv3_1"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 128
    pad: 0
    kernel_size: 1
    stride: 1
    bias_term: true
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}

layer {
  name: "conv3_1bn"
  type: "BatchNorm"
  bottom: "conv3_1"
  top: "conv3_1"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
}

layer {
  name: "conv3_1scale"
  type: "Scale"
  bottom: "conv3_1"
  top: "conv3_1"
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  param {
    lr_mult: 0.2
    decay_mult: 0
  }
  scale_param {
    bias_term: true
  }
}

layer {
  name: "relu3_1"
  type: "ReLU"
  bottom: "conv3_1"
  top: "conv3_1"
}
layer {
  name: "conv3_2"
  type: "Convolution"
  bottom: "conv3_1"
  top: "conv3_2"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 256
    pad: 1
    kernel_size: 3
    stride: 2
    bias_term: true
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}

layer {
  name: "conv3_2bn"
  type: "BatchNorm"
  bottom: "conv3_2"
  top: "conv3_2"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
}

layer {
  name: "conv3_2scale"
  type: "Scale"
  bottom: "conv3_2"
  top: "conv3_2"
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  param {
    lr_mult: 0.2
    decay_mult: 0
  }
  scale_param {
    bias_term: true
  }
}

layer {
  name: "relu3_2"
  type: "ReLU"
  bottom: "conv3_2"
  top: "conv3_2"
}

layer {
  name: "conv3_2_priorbox"
  type: "PriorBox"
  bottom: "conv3_2"
  bottom: "data"
  top: "conv3_2_priorbox"
  prior_box_param {
    fixed_size: 256
    density: 1
    step: 64
    variance: 0.1
    variance: 0.1
    variance: 0.2
    variance: 0.2
    offset: 0.5
  }
}

layer {
  name: "conv3_2_loc"
  type: "Convolution"
  bottom: "conv3_2"
  top: "conv3_2_loc"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 4
    pad: 1
    kernel_size: 3
    stride: 1
    bias_term: true
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "conv3_2_loc_perm"
  type: "Permute"
  bottom: "conv3_2_loc"
  top: "conv3_2_loc_perm"
  permute_param {
    order: 0
    order: 2
    order: 3
    order: 1
  }
}
layer {
  name: "conv3_2_loc_flat"
  type: "Flatten"
  bottom: "conv3_2_loc_perm"
  top: "conv3_2_loc_flat"
  flatten_param {
    axis: 1
  }
}

layer {
  name: "conv3_2_conf"
  type: "Convolution"
  bottom: "conv3_2"
  top: "conv3_2_conf"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 2
    pad: 1
    kernel_size: 3
    stride: 1
    bias_term: true
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "conv3_2_conf_perm"
  type: "Permute"
  bottom: "conv3_2_conf"
  top: "conv3_2_conf_perm"
  permute_param {
    order: 0
    order: 2
    order: 3
    order: 1
  }
}
layer {
  name: "conv3_2_conf_flat"
  type: "Flatten"
  bottom: "conv3_2_conf_perm"
  top: "conv3_2_conf_flat"
  flatten_param {
    axis: 1
  }
}

##################################conv4#########################
layer {
  name: "conv4_1"
  type: "Convolution"
  bottom: "conv3_2"
  top: "conv4_1"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 128
    pad: 0
    kernel_size: 1
    stride: 1
    bias_term: true
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}

layer {
  name: "conv4_1bn"
  type: "BatchNorm"
  bottom: "conv4_1"
  top: "conv4_1"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
}

layer {
  name: "conv4_1scale"
  type: "Scale"
  bottom: "conv4_1"
  top: "conv4_1"
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  param {
    lr_mult: 0.2
    decay_mult: 0
  }
  scale_param {
    bias_term: true
  }
}

layer {
  name: "relu4_1"
  type: "ReLU"
  bottom: "conv4_1"
  top: "conv4_1"
}
layer {
  name: "conv4_2"
  type: "Convolution"
  bottom: "conv4_1"
  top: "conv4_2"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 256
    pad: 1
    kernel_size: 3
    stride: 2
    bias_term: true
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}

layer {
  name: "conv4_2bn"
  type: "BatchNorm"
  bottom: "conv4_2"
  top: "conv4_2"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
}

layer {
  name: "conv4_2scale"
  type: "Scale"
  bottom: "conv4_2"
  top: "conv4_2"
  param {
    lr_mult: 0.1
    decay_mult: 0
  }
  param {
    lr_mult: 0.2
    decay_mult: 0
  }
  scale_param {
    bias_term: true
  }
}

layer {
  name: "relu4_2"
  type: "ReLU"
  bottom: "conv4_2"
  top: "conv4_2"
}

layer {
  name: "conv4_2_priorbox"
  type: "PriorBox"
  bottom: "conv4_2"
  bottom: "data"
  top: "conv4_2_priorbox"
  prior_box_param {
    fixed_size: 512
    density: 1
    step: 128
    variance: 0.1
    variance: 0.1
    variance: 0.2
    variance: 0.2
    offset: 0.5
  }
}

layer {
  name: "conv4_2_loc"
  type: "Convolution"
  bottom: "conv4_2"
  top: "conv4_2_loc"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 4
    pad: 1
    kernel_size: 3
    stride: 1
    bias_term: true
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "conv4_2_loc_perm"
  type: "Permute"
  bottom: "conv4_2_loc"
  top: "conv4_2_loc_perm"
  permute_param {
    order: 0
    order: 2
    order: 3
    order: 1
  }
}
layer {
  name: "conv4_2_loc_flat"
  type: "Flatten"
  bottom: "conv4_2_loc_perm"
  top: "conv4_2_loc_flat"
  flatten_param {
    axis: 1
  }
}

layer {
  name: "conv4_2_conf"
  type: "Convolution"
  bottom: "conv4_2"
  top: "conv4_2_conf"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 2
    pad: 1
    kernel_size: 3
    stride: 1
    bias_term: true
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "conv4_2_conf_perm"
  type: "Permute"
  bottom: "conv4_2_conf"
  top: "conv4_2_conf_perm"
  permute_param {
    order: 0
    order: 2
    order: 3
    order: 1
  }
}
layer {
  name: "conv4_2_conf_flat"
  type: "Flatten"
  bottom: "conv4_2_conf_perm"
  top: "conv4_2_conf_flat"
  flatten_param {
    axis: 1
  }
}

###############################################

layer {
  name: "mbox_priorbox"
  type: "Concat"
  bottom: "inception3_priorbox"
  bottom: "conv3_2_priorbox"
  bottom: "conv4_2_priorbox"
  top: "mbox_priorbox"
  concat_param {
    axis: 2
  }
}

layer {
  name: "mbox_loc"
  type: "Concat"
  bottom: "inception3_loc_flat"
  bottom: "conv3_2_loc_flat"
  bottom: "conv4_2_loc_flat"
  top: "mbox_loc"
  concat_param {
    axis: 1
  }
}


layer {
  name: "mbox_conf"
  type: "Concat"
  bottom: "inception3_conf_flat"
  bottom: "conv3_2_conf_flat"
  bottom: "conv4_2_conf_flat"
  top: "mbox_conf"
  concat_param {
    axis: 1
  }
}

layer {
  name: "mbox_conf_reshape"
  type: "Reshape"
  bottom: "mbox_conf"
  top: "mbox_conf_reshape"
  reshape_param {
    shape {
      dim: 0
      dim: -1
      dim: 2
    }
  }
}
layer {
  name: "mbox_conf_softmax"
  type: "Softmax"
  bottom: "mbox_conf_reshape"
  top: "mbox_conf_softmax"
  softmax_param {
    axis: 2
  }
}
layer {
  name: "mbox_conf_flatten"
  type: "Flatten"
  bottom: "mbox_conf_softmax"
  top: "mbox_conf_flatten"
  flatten_param {
    axis: 1
  }
}

#layer {
#  name: "detection_out"
#  type: "DetectionOutput"
#  bottom: "mbox_loc"
#  bottom: "mbox_conf_flatten"
#  bottom: "mbox_priorbox"
#  top: "detection_out"
#  include {
#    phase: TEST
#  }
#  detection_output_param {
#    num_classes: 2
#    share_location: true
#    background_label_id: 0
#    nms_param {
#      nms_threshold: 0.45
#      top_k: 100
#    }
#    code_type: CENTER_SIZE
#    keep_top_k: 100
#    confidence_threshold: 0.1
#  }
#}

layer {
  name: "detection_out"
  type: "DetectionOutput"
  bottom: "mbox_loc"
  bottom: "mbox_conf_flatten"
  bottom: "mbox_priorbox"
  top: "detection_out"
  include {
    phase: TEST
  }
  detection_output_param {
    num_classes: 2
    share_location: true
    background_label_id: 0
    nms_param {
      nms_threshold: 0.3
      top_k: 400
    }
    code_type: CENTER_SIZE
    keep_top_k: 200
    confidence_threshold: 0.1
  }
}