train_val.prototxt

name: "CaffeNet"
# ************************
# ***** DATA LAYER *******
# ************************
layer {
  name: "data"
  type: "Data"
  top: "data"
  include {    phase: TRAIN  }
  transform_param {
    mirror: 1  # 1 = on, 0 = off
    crop_size: 180
  }
  data_param {
    source: "./data/caffe-train-lmdb"
    batch_size: 80
    backend: LMDB
  }
}
layer {
  name: "data"
  type: "Data"
  top: "data"
  include {    phase: TEST  }
  transform_param {
    mirror: 0  # 1 = on, 0 = off
    crop_size: 256
  }
  data_param {
    source: "./data/caffe-val-lmdb"
    batch_size: 10
    backend: LMDB
  }
}

layer { # Convert to lab
  name: "img_lab"
  type: "ColorConv"
  # bottom: "img"
  bottom: "data"
  top: "img_lab"
  propagate_down: false
  color_conv_param {
    input: 0 # BGR
    output: 3 # Lab
  }
}
layer {
  name: "img_slice"
  type: "Slice"
  bottom: "img_lab"
  top: "img_l" # [0,100]
  top: "data_ab" # [-110,110]
  propagate_down: false
  slice_param {
    axis: 1
    slice_point: 1
  }
}
layer { # 0-center lightness channel
  name: "data_l"
  type: "Convolution"
  bottom: "img_l"
  top: "data_l" # [-50,50]
  propagate_down: false
  param {lr_mult: 0 decay_mult: 0}
  param {lr_mult: 0 decay_mult: 0}
  convolution_param {
    kernel_size: 1
    num_output: 1
    weight_filler { type: 'constant' value: 1}
    bias_filler { type: 'constant' value: -50}
  }
}
# ****************************
# ***** PROCESS LABELS *******
# ****************************
layer { # subsample ab
  name: 'data_ab_ss'
  type: 'Convolution'
  bottom: "data_ab"
  top: "data_ab_ss" # quantized gt colors
  param {lr_mult: 0 decay_mult: 0}
  param {lr_mult: 0 decay_mult: 0}
  convolution_param {
    num_output: 2
    kernel_size: 1
    stride: 8
    group: 2
    weight_filler { type: 'constant' value: 1 }
  }
}
layer { # encode
  name: 'ab_enc'
  type: 'Python'
  bottom: "data_ab_ss"
  top: "gt_ab_ss" # quantized gt colors
  python_param {
    module: 'caffe_traininglayers'
    layer: 'NNEncLayer'
  }
}
layer { # compute gray mask
  name: 'ab_graymask'
  type: 'Python'
  bottom: "data_ab_ss"
  top: "nongray_mask"
  python_param {
    module: 'caffe_traininglayers'
    layer: 'NonGrayMaskLayer'
  }
}
layer { # compute prior boost
  type: 'Python'
  name: 'ab_pb'
  bottom: "gt_ab_ss"
  top: "prior_boost"
  python_param {
    module: 'caffe_traininglayers'
    layer: 'PriorBoostLayer'
  }
}
layer { # multiply nongray mask and prior boost
  type: 'Eltwise'
  name: 'pb_nongray'
  bottom: "prior_boost"
  bottom: "nongray_mask"
  top: "prior_boost_nongray" # prior boost to use 
  eltwise_param {    operation: 0  }
}
# *****************
# ***** Conv1 *****
# *****************
layer {
  name: "conv1"
  type: "Convolution"
  bottom: "data_l"
  top: "conv1"
  param {    lr_mult: 1    decay_mult: 1  }
  param {    lr_mult: 2    decay_mult: 0  }
  convolution_param {
    num_output: 96
    kernel_size: 11
    pad: 5
    stride: 4
  }
}
layer {
  name: "conv1norm"
  type: "BatchNorm"
  bottom: "conv1"
  top: "conv1norm"
  batch_norm_param{ }
  param {lr_mult: 0 decay_mult: 0}
  param {lr_mult: 0 decay_mult: 0}
  param {lr_mult: 0 decay_mult: 0}
}
layer {
  name: "relu1"
  type: "ReLU"
  bottom: "conv1norm"
  top: "conv1norm"
}
layer {
  name: "pool1"
  type: "Pooling"
  bottom: "conv1norm"
  top: "pool1"
  pooling_param {
    pool: MAX
    kernel_size: 3
    stride: 2
    pad: 1
  }
}
layer {
  name: "conv2"
  type: "Convolution"
  bottom: "pool1"
  top: "conv2"
  param {    lr_mult: 1    decay_mult: 1  }
  param {    lr_mult: 2    decay_mult: 0  }
  convolution_param {
    num_output: 256
    pad: 2
    kernel_size: 5
    group: 2
  }
}
layer {
  name: "conv2norm"
  type: "BatchNorm"
  bottom: "conv2"
  top: "conv2norm"
  batch_norm_param{ }
  param {lr_mult: 0 decay_mult: 0}
  param {lr_mult: 0 decay_mult: 0}
  param {lr_mult: 0 decay_mult: 0}
}
layer {
  name: "relu2"
  type: "ReLU"
  bottom: "conv2norm"
  top: "conv2norm"
}
layer {
  name: "pool2"
  type: "Pooling"
  # bottom: "conv2"
  bottom: "conv2norm"
  top: "pool2"
  pooling_param {
    pool: MAX
    kernel_size: 3
    stride: 2
    pad: 1
  }
}
layer {
  name: "conv3"
  type: "Convolution"
  bottom: "pool2"
  top: "conv3"
  param {    lr_mult: 1    decay_mult: 1  }
  param {    lr_mult: 2    decay_mult: 0  }
  convolution_param {
    num_output: 384
    dilation: 1
    pad: 1
    kernel_size: 3
  }
}
layer {
  name: "conv3norm"
  type: "BatchNorm"
  bottom: "conv3"
  top: "conv3norm"
  batch_norm_param{ }
  param {lr_mult: 0 decay_mult: 0}
  param {lr_mult: 0 decay_mult: 0}
  param {lr_mult: 0 decay_mult: 0}
}
layer {
  name: "relu3"
  type: "ReLU"
  bottom: "conv3norm"
  top: "conv3norm"
}
layer {
  name: "conv4"
  type: "Convolution"
  bottom: "conv3norm"
  top: "conv4"
  param {    lr_mult: 1    decay_mult: 1  }
  param {    lr_mult: 2    decay_mult: 0  }
  convolution_param {
    num_output: 384
    dilation: 1
    pad: 1
    kernel_size: 3
    group: 2
  }
}
layer {
  name: "conv4norm"
  type: "BatchNorm"
  bottom: "conv4"
  top: "conv4norm"
  batch_norm_param{ }
  param {lr_mult: 0 decay_mult: 0}
  param {lr_mult: 0 decay_mult: 0}
  param {lr_mult: 0 decay_mult: 0}
}
layer {
  name: "relu4"
  type: "ReLU"
  bottom: "conv4norm"
  top: "conv4norm"
}
layer {
  name: "conv5"
  type: "Convolution"
  bottom: "conv4norm"
  top: "conv5"
  param {    lr_mult: 1    decay_mult: 1  }
  param {    lr_mult: 2    decay_mult: 0  }
  convolution_param {
    num_output: 256
    dilation: 1
    pad: 1
    kernel_size: 3
    group: 2
  }
}
layer {
  name: "conv5norm"
  type: "BatchNorm"
  bottom: "conv5"
  top: "conv5norm"
  batch_norm_param{ }
  param {lr_mult: 0 decay_mult: 0}
  param {lr_mult: 0 decay_mult: 0}
  param {lr_mult: 0 decay_mult: 0}
}
layer {
  name: "relu5"
  type: "ReLU"
  bottom: "conv5norm"
  top: "conv5norm"
}
layer {
  name: "pool5"
  type: "Pooling"
  bottom: "conv5norm"
  top: "pool5"
  pooling_param {
    pool: MAX
    kernel_size: 3
    pad: 1
    stride: 1
  }
}
layer {
  name: "fc6"
  type: "Convolution"
  bottom: "pool5"
  top: "fc6"
  param {    lr_mult: 1    decay_mult: 1  }
  param {    lr_mult: 2    decay_mult: 0  }
  convolution_param {
    kernel_size: 6
    dilation: 2
    pad: 5
    stride: 1
    num_output: 4096
    weight_filler {
      type: "gaussian"
      std: 0.005
    }
    bias_filler {
      type: "constant"
      value: 1
    }
  }
}
layer {
  name: "fc6norm"
  type: "BatchNorm"
  bottom: "fc6"
  top: "fc6norm"
  batch_norm_param{ }
  param {lr_mult: 0 decay_mult: 0}
  param {lr_mult: 0 decay_mult: 0}
  param {lr_mult: 0 decay_mult: 0}
}
layer {
  name: "relu6"
  type: "ReLU"
  bottom: "fc6norm"
  top: "fc6norm"
}
layer {
  name: "fc7"
  type: "Convolution"
  bottom: "fc6norm"
  top: "fc7"
  param {    lr_mult: 1    decay_mult: 1  }
  param {    lr_mult: 2    decay_mult: 0  }
  convolution_param {
    kernel_size: 1
    stride: 1
    num_output: 4096
    weight_filler {
      type: "gaussian"
      std: 0.005
    }
    bias_filler {
      type: "constant"
      value: 1
    }
  }
}
layer {
  name: "fc7norm"
  type: "BatchNorm"
  bottom: "fc7"
  top: "fc7norm"
  batch_norm_param{ }
  param {lr_mult: 0 decay_mult: 0}
  param {lr_mult: 0 decay_mult: 0}
  param {lr_mult: 0 decay_mult: 0}
}
layer {
  name: "relu7"
  type: "ReLU"
  bottom: "fc7norm"
  top: "fc7norm"
}
layer {
  name: "fc8"
  type: "Convolution"
  bottom: "fc7norm"
  top: "fc8"
  param {    lr_mult: 1    decay_mult: 1  }
  param {    lr_mult: 2    decay_mult: 0  }
  convolution_param {
    kernel_size: 1
    num_output: 313
  }
}
# ******************************
# ***** POSTPROCESS LABELS *****
# ******************************
layer {
  name: "fc8_upsample"
  type: "Resize"
  bottom: "fc8"
  bottom: "gt_ab_ss"
  top: "fc8_us"
}
layer {
  name: "PriorBoost8"
  type: "Python"
  bottom: "fc8_us"
  bottom: "prior_boost_nongray"
  top: "fc8_boost"
  python_param {
    module: 'caffe_traininglayers'
    layer: 'ClassRebalanceMultLayer'
  }
}
layer {
  name: "Softmax"
  type: "SoftmaxCrossEntropyLoss"
  bottom: "fc8_boost"
  bottom: "gt_ab_ss"
  top: "loss"
  loss_weight: 1.
  loss_param {    normalize: True  }
}