# VGG 16-layer network convolutional finetuning
# Network modified to have smaller receptive field (128 pixels)
# and smaller stride (8 pixels) when run in convolutional mode.
#
# In this model we also change max pooling size in the first 4 layer
# from 2 to 3 while retaining stride = 2
# which makes it easier to exactly align responses at different layer.
#
name: "deeplab_msc_largeFOV"

layer {
  name: "data"
  type: "ImageSegData"
  top: "data"
  top: "label"
  top: "data_dim"
  image_data_param {
    root_folder: "/home/m62tang/rloss/data/pascal_scribble/"
    source: "pascal_scribble/list/val.txt"
    label_type: PIXEL
    batch_size: 1
  }
  transform_param {
    mean_value: 104.008
    mean_value: 116.669
    mean_value: 122.675
    crop_size: 513
    mirror: false
  }
  include: { phase: TEST }
}

# BEG (0) Direct path to classifier
layer {
  bottom: "data"
  top: "data_conv"
  name: "data_conv"
  type: "Convolution"
  convolution_param {
    num_output: 128
    kernel_size: 3
    stride: 1
    pad: 1
  }
}
layer {
  bottom: "data_conv"
  top: "data_conv"
  name: "relu_data_conv"
  type: "ReLU"
}
layer {
  bottom: "data_conv"
  top: "data_conv"
  name: "drop_data_conv"
  type: "Dropout"
  dropout_param {
    dropout_ratio: 0.5
  }
}
layer {
  bottom: "data_conv"
  top: "data_fc"
  name: "data_fc"
  type: "Convolution"
  convolution_param {
    num_output: 128
    kernel_size: 1
    stride: 1
  }
}
layer {
  bottom: "data_fc"
  top: "data_fc"
  name: "relu_data_fc"
  type: "ReLU"
}
layer {
  bottom: "data_fc"
  top: "data_fc"
  name: "drop_data_fc"
  type: "Dropout"
  dropout_param {
    dropout_ratio: 0.5
  }
}
layer {
  bottom: "data_fc"
  top: "data_ms"
  name: "data_ms"
  type: "Convolution"
  convolution_param {
    num_output: 21
    kernel_size: 1
  }
}
# END (0) Direct path to classifier

### NETWORK ###

layer {
  bottom: "data"
  top: "conv1_1"
  name: "conv1_1"
  type: "Convolution"
  convolution_param {
    num_output: 64
    pad: 1
    kernel_size: 3
  }
}
layer {
  bottom: "conv1_1"
  top: "conv1_1"
  name: "relu1_1"
  type: "ReLU"
}
layer {
  bottom: "conv1_1"
  top: "conv1_2"
  name: "conv1_2"
  type: "Convolution"
  convolution_param {
    num_output: 64
    pad: 1
    kernel_size: 3
  }
}
layer {
  bottom: "conv1_2"
  top: "conv1_2"
  name: "relu1_2"
  type: "ReLU"
}
layer {
  bottom: "conv1_2"
  top: "pool1"
  name: "pool1"
  type: "Pooling"
  pooling_param {
    pool: MAX
    kernel_size: 3
    stride: 2
    pad: 1
  }
}
# BEG (1) Direct path to classifier
layer {
  bottom: "pool1"
  top: "pool1_conv"
  name: "pool1_conv"
  type: "Convolution"
  convolution_param {
    num_output: 128
    kernel_size: 3
    stride:1
    pad: 1
  }
}
layer {
  bottom: "pool1_conv"
  top: "pool1_conv"
  name: "relu_pool1_conv"
  type: "ReLU"
}
layer {
  bottom: "pool1_conv"
  top: "pool1_conv"
  name: "drop_pool1_conv"
  type: "Dropout"
  dropout_param {
    dropout_ratio: 0.5
  }
}
layer {
  bottom: "pool1_conv"
  top: "pool1_fc"
  name: "pool1_fc"
  type: "Convolution"
  convolution_param {
    num_output: 128
    kernel_size: 1
  }
}
layer {
  bottom: "pool1_fc"
  top: "pool1_fc"
  name: "relu_pool1_fc"
  type: "ReLU"
}
layer {
  bottom: "pool1_fc"
  top: "pool1_fc"
  name: "drop_pool1_fc"
  type: "Dropout"
  dropout_param {
    dropout_ratio: 0.5
  }
}
layer {
  bottom: "pool1_fc"
  top: "pool1_ms"
  name: "pool1_ms"
  type: "Convolution"
  convolution_param {
    num_output: 21
    kernel_size: 1
  }
}
# END (1) Direct path to classifier
layer {
  bottom: "pool1"
  top: "conv2_1"
  name: "conv2_1"
  type: "Convolution"
  convolution_param {
    num_output: 128
    pad: 1
    kernel_size: 3
  }
}
layer {
  bottom: "conv2_1"
  top: "conv2_1"
  name: "relu2_1"
  type: "ReLU"
}
layer {
  bottom: "conv2_1"
  top: "conv2_2"
  name: "conv2_2"
  type: "Convolution"
  convolution_param {
    num_output: 128
    pad: 1
    kernel_size: 3
  }
}
layer {
  bottom: "conv2_2"
  top: "conv2_2"
  name: "relu2_2"
  type: "ReLU"
}
layer {
  bottom: "conv2_2"
  top: "pool2"
  name: "pool2"
  type: "Pooling"
  pooling_param {
    pool: MAX
    kernel_size: 3
    stride: 2
    pad: 1
  }
}
# BEG (2) Direct path to classifier
layer {
  bottom: "pool2"
  top: "pool2_conv"
  name: "pool2_conv"
  type: "Convolution"
  convolution_param {
    num_output: 128
    kernel_size: 3
    stride: 1
    pad: 1
  }
}
layer {
  bottom: "pool2_conv"
  top: "pool2_conv"
  name: "relu_pool2_conv"
  type: "ReLU"
}
layer {
  bottom: "pool2_conv"
  top: "pool2_conv"
  name: "drop_pool2_conv"
  type: "Dropout"
  dropout_param {
    dropout_ratio: 0.5
  }
}
layer {
  bottom: "pool2_conv"
  top: "pool2_fc"
  name: "pool2_fc"
  type: "Convolution"
  convolution_param {
    num_output: 128
    kernel_size: 1
  }
}
layer {
  bottom: "pool2_fc"
  top: "pool2_fc"
  name: "relu_pool2_fc"
  type: "ReLU"
}
layer {
  bottom: "pool2_fc"
  top: "pool2_fc"
  name: "drop_pool2_fc"
  type: "Dropout"
  dropout_param {
    dropout_ratio: 0.5
  }
}
layer {
  bottom: "pool2_fc"
  top: "pool2_ms"
  name: "pool2_ms"
  type: "Convolution"
  convolution_param {
    num_output: 21
    kernel_size: 1
  }
}
# END (2) Direct path to classifier
layer {
  bottom: "pool2"
  top: "conv3_1"
  name: "conv3_1"
  type: "Convolution"
  convolution_param {
    num_output: 256
    pad: 1
    kernel_size: 3
  }
}
layer {
  bottom: "conv3_1"
  top: "conv3_1"
  name: "relu3_1"
  type: "ReLU"
}
layer {
  bottom: "conv3_1"
  top: "conv3_2"
  name: "conv3_2"
  type: "Convolution"
  convolution_param {
    num_output: 256
    pad: 1
    kernel_size: 3
  }
}
layer {
  bottom: "conv3_2"
  top: "conv3_2"
  name: "relu3_2"
  type: "ReLU"
}
layer {
  bottom: "conv3_2"
  top: "conv3_3"
  name: "conv3_3"
  type: "Convolution"
  convolution_param {
    num_output: 256
    pad: 1
    kernel_size: 3
  }
}
layer {
  bottom: "conv3_3"
  top: "conv3_3"
  name: "relu3_3"
  type: "ReLU"
}
layer {
  bottom: "conv3_3"
  top: "pool3"
  name: "pool3"
  type: "Pooling"
  pooling_param {
    pool: MAX
    kernel_size: 3
    stride: 2
    pad: 1
  }
}
# BEG (3) Direct path to classifier
layer {
  bottom: "pool3"
  top: "pool3_conv"
  name: "pool3_conv"
  type: "Convolution"
  convolution_param {
    num_output: 128
    kernel_size: 3
    stride: 1
    pad: 1
  }
}
layer {
  bottom: "pool3_conv"
  top: "pool3_conv"
  name: "relu_pool3_conv"
  type: "ReLU"
}
layer {
  bottom: "pool3_conv"
  top: "pool3_conv"
  name: "drop_pool3_conv"
  type: "Dropout"
  dropout_param {
    dropout_ratio: 0.5
  }
}
layer {
  bottom: "pool3_conv"
  top: "pool3_fc"
  name: "pool3_fc"
  type: "Convolution"
  convolution_param {
    num_output: 128
    kernel_size: 1
    stride: 1
  }
}
layer {
  bottom: "pool3_fc"
  top: "pool3_fc"
  name: "relu_pool3_fc"
  type: "ReLU"
}
layer {
  bottom: "pool3_fc"
  top: "pool3_fc"
  name: "drop_pool3_fc"
  type: "Dropout"
  dropout_param {
    dropout_ratio: 0.5
  }
}
layer {
  bottom: "pool3_fc"
  top: "pool3_ms"
  name: "pool3_ms"
  type: "Convolution"
  convolution_param {
    num_output: 21
    kernel_size: 1
  }
}
# END (3) Direct path to classifier
layer {
  bottom: "pool3"
  top: "conv4_1"
  name: "conv4_1"
  type: "Convolution"
  convolution_param {
    num_output: 512
    pad: 1
    kernel_size: 3
  }
}
layer {
  bottom: "conv4_1"
  top: "conv4_1"
  name: "relu4_1"
  type: "ReLU"
}
layer {
  bottom: "conv4_1"
  top: "conv4_2"
  name: "conv4_2"
  type: "Convolution"
  convolution_param {
    num_output: 512
    pad: 1
    kernel_size: 3
  }
}
layer {
  bottom: "conv4_2"
  top: "conv4_2"
  name: "relu4_2"
  type: "ReLU"
}
layer {
  bottom: "conv4_2"
  top: "conv4_3"
  name: "conv4_3"
  type: "Convolution"
  convolution_param {
    num_output: 512
    pad: 1
    kernel_size: 3
  }
}
layer {
  bottom: "conv4_3"
  top: "conv4_3"
  name: "relu4_3"
  type: "ReLU"
}
layer {
  bottom: "conv4_3"
  top: "pool4"
  name: "pool4"
  type: "Pooling"
  pooling_param {
    pool: MAX
    kernel_size: 3
    pad: 1
    #stride: 2
    stride: 1
  }
}
# BEG (4) Direct path to classifier
layer {
  bottom: "pool4"
  top: "pool4_conv"
  name: "pool4_conv"
  type: "Convolution"
  convolution_param {
    num_output: 128
    kernel_size: 3
    stride: 1
    pad: 1
  }
}
layer {
  bottom: "pool4_conv"
  top: "pool4_conv"
  name: "relu_pool4_conv"
  type: "ReLU"
}
layer {
  bottom: "pool4_conv"
  top: "pool4_conv"
  name: "drop_pool4_conv"
  type: "Dropout"
  dropout_param {
    dropout_ratio: 0.5
  }
}
layer {
  bottom: "pool4_conv"
  top: "pool4_fc"
  name: "pool4_fc"
  type: "Convolution"
  convolution_param {
    num_output: 128
    kernel_size: 1
    stride: 1
    weight_filler {
      type: "gaussian"
      std: 0.005
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  bottom: "pool4_fc"
  top: "pool4_fc"
  name: "relu_pool4_fc"
  type: "ReLU"
}
layer {
  bottom: "pool4_fc"
  top: "pool4_fc"
  name: "drop_pool4_fc"
  type: "Dropout"
  dropout_param {
    dropout_ratio: 0.5
  }
}
layer {
  bottom: "pool4_fc"
  top: "pool4_ms"
  name: "pool4_ms"
  type: "Convolution"
  convolution_param {
    num_output: 21
    kernel_size: 1
  }
}
# END (4) Direct path to classifier

layer {
  bottom: "pool4"
  top: "conv5_1"
  name: "conv5_1"
  type: "Convolution"
  convolution_param {
    num_output: 512
    #pad: 1
    pad: 2
    dilation: 2
    kernel_size: 3
  }
}
layer {
  bottom: "conv5_1"
  top: "conv5_1"
  name: "relu5_1"
  type: "ReLU"
}
layer {
  bottom: "conv5_1"
  top: "conv5_2"
  name: "conv5_2"
  type: "Convolution"
  convolution_param {
    num_output: 512
    #pad: 1
    pad: 2
    dilation: 2
    kernel_size: 3
  }
}
layer {
  bottom: "conv5_2"
  top: "conv5_2"
  name: "relu5_2"
  type: "ReLU"
}
layer {
  bottom: "conv5_2"
  top: "conv5_3"
  name: "conv5_3"
  type: "Convolution"
  convolution_param {
    num_output: 512
    #pad: 1
    pad: 2
    dilation: 2
    kernel_size: 3
  }
}
layer {
  bottom: "conv5_3"
  top: "conv5_3"
  name: "relu5_3"
  type: "ReLU"
}
layer {
  bottom: "conv5_3"
  top: "pool5"
  name: "pool5"
  type: "Pooling"
  pooling_param {
    pool: MAX
    #kernel_size: 2
    #stride: 2
    kernel_size: 3
    stride: 1
    pad: 1
  }
}

layer {
  bottom: "pool5"
  top: "pool5a"
  name: "pool5a"
  type: "Pooling"
  pooling_param {
    pool: AVE
    kernel_size: 3
    stride: 1
    pad: 1
  }
}

layer {
  bottom: "pool5a"
  top: "fc6"
  name: "fc6"
  type: "Convolution"
  #strict_dim: false
  convolution_param {
    num_output: 1024
    pad: 12
    dilation: 12
    kernel_size: 3
  }
}
layer {
  bottom: "fc6"
  top: "fc6"
  name: "relu6"
  type: "ReLU"
}
layer {
  bottom: "fc6"
  top: "fc6"
  name: "drop6"
  type: "Dropout"
  dropout_param {
    dropout_ratio: 0.5
  }
}
layer {
  bottom: "fc6"
  top: "fc7"
  name: "fc7"
  type: "Convolution"
  #strict_dim: false
  convolution_param {
    num_output: 1024
    kernel_size: 1
  }
}
layer {
  bottom: "fc7"
  top: "fc7"
  name: "relu7"
  type: "ReLU"
}
layer {
  bottom: "fc7"
  top: "fc7"
  name: "drop7"
  type: "Dropout"
  dropout_param {
    dropout_ratio: 0.5
  }
}
layer {
  bottom: "fc7"
  top: "fc8_pascal_scribble"
  name: "fc8_pascal_scribble"
  type: "Convolution"
  #strict_dim: false
  convolution_param {
    num_output: 21
    kernel_size: 1
  }
}
# Upsampling and fusion
layer {
  bottom: "pool1_ms"
  top: "pool1_interp"
  name: "pool1_interp"
  type: "Interp"
  interp_param {
    zoom_factor: 2
  }
}
layer {
  bottom: "pool2_ms"
  top: "pool2_interp"
  name: "pool2_interp"
  type: "Interp"
  interp_param {
    zoom_factor: 4
  }
}
layer {
  bottom: "pool3_ms"
  top: "pool3_interp"
  name: "pool3_interp"
  type: "Interp"
  interp_param {
    zoom_factor: 8
  }
}
layer {
  bottom: "pool4_ms"
  top: "pool4_interp"
  name: "pool4_interp"
  type: "Interp"
  interp_param {
    zoom_factor: 8
  }
}
layer {
  bottom: "fc8_pascal_scribble"
  top: "fc8_interp"
  name: "fc8_interp"
  type: "Interp"
  interp_param {
    zoom_factor: 8
  }
}

# Fusion layer
layer {
  bottom: "data_ms"
  bottom: "pool1_interp"
  bottom: "pool2_interp"
  bottom: "pool3_interp"
  bottom: "pool4_interp"
  bottom: "fc8_interp"
  top: "fc_fusion"
  name: "fc_fusion"
  type: "Eltwise"
  eltwise_param {
    operation: SUM
  }
}

# JAY EDIT THIS LAYER
#layer {
#  bottom: "fc8_interp"
#  bottom: "data_dim"
#  bottom: "data"
#  top: "crf_inf"
#  name: "crf"
#  type: DENSE_CRF
#  dense_crf_param { 
#    max_iter: 10
#    pos_w: 3
#    pos_xy_std: 3
#    bi_w: 5
#    bi_xy_std: 50
#    bi_rgb_std: 10
#  }
#  include: { phase: TEST }
#}

#layer {
#  name: "crf_mat"
#  type: MAT_WRITE
#  bottom: "crf_inf"
#  mat_write_param {
#    prefix: "pascal_scribble/features/deeplab_msc_largeFOV/val/crf/"
#    source: "pascal_scribble/list/val_id.txt"
#    strip: 0
#    period: 1
#  }
#  include: { phase: TEST }
#}

# layer {
#   bottom: "label"
#   name: "silence"
#   type: SILENCE
#   include: { phase: TEST }
# }

layer {
  name: "accuracy"
  type: "SegAccuracy"
  bottom: "fc_fusion"
  bottom: "label"
  top: "accuracy"
  seg_accuracy_param {
    ignore_label: 255
    reset: false
  } 
}


########### DENSE CRF ############
#layer {
#  bottom: "fc_fusion"
#  bottom: "data_dim"
#  bottom: "data"
#  top: "crf_inf"
#  name: "crf"
#  type: "DenseCRF"
#  dense_crf_param {
#    max_iter: 10
#    pos_w: 3
#    pos_xy_std: 3
#    bi_w: 4
#    bi_xy_std: 83
#    bi_rgb_std: 5
#  }
#}

#layer {
#  name: "accuracyCRF"
#  type: "SegAccuracy"
#  bottom: "crf_inf"
#  bottom: "label"
#  top: "accuracyCRF"
#  seg_accuracy_param {
#    ignore_label: 255
#    reset: false
# } 
#}

#layer {
#  name: "crf_mat"
#  type: "MatWrite"
#  bottom: "crf_inf"
#  include {
#    phase: TEST
#  }
#  mat_write_param {
#    prefix: "pascal_scribble/features/deeplab_msc_largeFOV/val/crf/"
#    source: "pascal_scribble/list/val_id.txt"
#    strip: 0
#    period: 1
#  }
#}