name: "pointnet_cls_basic"

# ---------------------- data layer ---------------
#layer {
#  name: "data"
#  type: "HDF5Data"
#  top: "data"
#  top: "label"
#  hdf5_data_param {
#    source: "data/modelnet40_ply_hdf5_2048/train_files.txt"
#    batch_size: 32
#  }
#  include: { phase: TRAIN }
#}
#layer {
 # name: "data"
 # type: "HDF5Data"
 # top: "data"
 # top: "label"
 # hdf5_data_param {
 #   source: "data/modelnet40_ply_hdf5_2048/test_files.txt"
 #   batch_size: 32
 # }
 # include: { phase: TEST }
#}
layer {
    name: "reshape"
    type: "Reshape"
    bottom: "data"
    top: "data_reshape"
    reshape_param {
      shape {
        dim: 0  # copy the dimension from below
        dim: 1
        dim: -1
        dim: 3 # infer it from the other dimensions
      }
    }
}
# -------------------- input tansform net ---------------------
layer {
  name: "conv1_tnet1"
  type: "Convolution"
  bottom: "data_reshape"
  top: "conv1_tnet1"
  param {
    lr_mult: 1
  }
  param {
    lr_mult: 2
  }
  convolution_param {
    num_output: 64
    pad: 0
    kernel_w: 3
    kernel_h: 1
    stride: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "bn1_tnet1"
  type: "BatchNorm"
  bottom: "conv1_tnet1"
  top: "conv1_tnet1"
}
layer {
    bottom: "conv1_tnet1"
    top: "conv1_tnet1"
    name: "scale1_tnet1"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}
layer {
  name: "relu1_tnet1"
  type: "ReLU"
  bottom: "conv1_tnet1"
  top: "conv1_tnet1"
}

layer {
  name: "conv2_tnet1"
  type: "Convolution"
  bottom: "conv1_tnet1"
  top: "conv2_tnet1"
  param {
    lr_mult: 1
  }
  param {
    lr_mult: 2
  }
  convolution_param {
    num_output: 128
    pad: 0
    kernel_w: 1
    kernel_h: 1
    stride: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "bn2_tnet1"
  type: "BatchNorm"
  bottom: "conv2_tnet1"
  top: "conv2_tnet1"
}
layer {
    bottom: "conv2_tnet1"
    top: "conv2_tnet1"
    name: "scale2_tnet1"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}
layer {
  name: "relu2_tnet1"
  type: "ReLU"
  bottom: "conv2_tnet1"
  top: "conv2_tnet1"
}

layer {
  name: "conv3_tnet1"
  type: "Convolution"
  bottom: "conv2_tnet1"
  top: "conv3_tnet1"
  param {
    lr_mult: 1
  }
  param {
    lr_mult: 2
  }
  convolution_param {
    num_output: 1024
    pad: 0
    kernel_w: 1
    kernel_h: 1
    stride: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "bn3_tnet1"
  type: "BatchNorm"
  bottom: "conv3_tnet1"
  top: "conv3_tnet1"
}
layer {
    bottom: "conv3_tnet1"
    top: "conv3_tnet1"
    name: "scale3_tnet1"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}
layer {
  name: "relu3_tnet1"
  type: "ReLU"
  bottom: "conv3_tnet1"
  top: "conv3_tnet1"
}
layer {
  name: "pool_tnet1"
  type: "Pooling"
  bottom: "conv3_tnet1"
  top: "global_feat_tnet1"
  pooling_param {
    pool: MAX
    pad: 0
    kernel_h: 2048
    kernel_w: 1
    stride: 1
  }
}
layer {
  name: "fc1_tnet1"
  type: "InnerProduct"
  bottom: "global_feat_tnet1"
  top: "fc1_tnet1"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 1
    decay_mult: 1
  }
  inner_product_param {
    num_output: 512
    weight_filler {
      type: "gaussian"
      std: 0.001
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "bn6_tnet1"
  type: "BatchNorm"
  bottom: "fc1_tnet1"
  top: "fc1_tnet1"
}
layer {
    bottom: "fc1_tnet1"
    top: "fc1_tnet1"
    name: "scale6_tnet1"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}
layer {
  name: "relu6_tnet1"
  type: "ReLU"
  bottom: "fc1_tnet1"
  top: "fc1_tnet1"
}

layer {
  name: "fc2_tnet1"
  type: "InnerProduct"
  bottom: "fc1_tnet1"
  top: "fc2_tnet1"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 1
    decay_mult: 1
  }
  inner_product_param {
    num_output: 256
    weight_filler {
      type: "gaussian"
      std: 0.001
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "bn7_tnet1"
  type: "BatchNorm"
  bottom: "fc2_tnet1"
  top: "fc2_tnet1"
}
layer {
    bottom: "fc2_tnet1"
    top: "fc2_tnet1"
    name: "scale7_tnet1"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}
layer {
  name: "relu7_tnet1"
  type: "ReLU"
  bottom: "fc2_tnet1"
  top: "fc2_tnet1"
}
layer {
  name: "fc3_tnet1"
  type: "InnerProduct"
  bottom: "fc2_tnet1"
  top: "fc3_tnet1"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 1
    decay_mult: 1
  }
  inner_product_param {
    num_output: 9
    weight_filler {
      type: "constant"
      value: 0
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
    name: "reshape_tnet1"
    type: "Reshape"
    bottom: "fc3_tnet1"
    top: "fc3_tnet1_reshape"
    reshape_param {
      shape {
		dim: -1
        dim: 3
        dim: 3
      }
    }
}
#layer {
#  name: "eye_tnet1"
#  type: "Python"
#  bottom: "fc3_tnet1"
#  top: "eye_tnet1"
#  python_param {
 #   module: "eye_matrix_layer"
 #   layer: "EyeMatrixLayer"
 #   param_str: "{\'K\': 3}"
 # }
#}
layer {
  name: "eltwise_sum_tnet1"
  type: "Eltwise"
  bottom: "fc3_tnet1_reshape"
 # bottom: "eye_tnet1"
  top: "transform1"
  eltwise_param { operation: SUM }
}
# -------------------- input transform ---------------------
#layer {
#  name: "matmul_input"
#  type: "MatrixMultiplication"
#  bottom: "data" # BxNx3
#  bottom: "transform1" # 3x3
#  top: "data_transform1"  # 10x1x1 # BxNx3
#}
layer {
    name: "reshape_input_data"
    type: "Reshape"
    bottom: "data_transform1"
    top: "data_transform1_reshape"
    reshape_param {
      shape {
        dim: 0  # copy the dimension from below
        dim: 1
        dim: -1
        dim: 3 # infer it from the other dimensions
      }
    }
}
# -------------------- mlp ---------------------
layer {
  name: "conv1"
  type: "Convolution"
  bottom: "data_transform1_reshape"
  top: "conv1"
  param {
    lr_mult: 1
  }
  param {
    lr_mult: 2
  }
  convolution_param {
    num_output: 64
    pad: 0
    kernel_w: 3
    kernel_h: 1
    stride: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "bn1"
  type: "BatchNorm"
  bottom: "conv1"
  top: "conv1"
}
layer {
    bottom: "conv1"
    top: "conv1"
    name: "scale1"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}
layer {
  name: "relu1"
  type: "ReLU"
  bottom: "conv1"
  top: "conv1"
}

layer {
  name: "conv2"
  type: "Convolution"
  bottom: "conv1"
  top: "conv2"
  param {
    lr_mult: 1
  }
  param {
    lr_mult: 2
  }
  convolution_param {
    num_output: 64
    pad: 0
    kernel_w: 1
    kernel_h: 1
    stride: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "bn2"
  type: "BatchNorm"
  bottom: "conv2"
  top: "conv2"
}
layer {
    bottom: "conv2"
    top: "conv2"
    name: "scale2"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}
layer {
  name: "relu2"
  type: "ReLU"
  bottom: "conv2"
  top: "conv2"
}


# -------------------- feat tansform net ---------------------
layer {
  name: "conv1_tnet2"
  type: "Convolution"
  bottom: "conv2"
  top: "conv1_tnet2"
  param {
    lr_mult: 1
  }
  param {
    lr_mult: 2
  }
  convolution_param {
    num_output: 64
    pad: 0
    kernel_w: 1
    kernel_h: 1
    stride: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "bn1_tnet2"
  type: "BatchNorm"
  bottom: "conv1_tnet2"
  top: "conv1_tnet2"
}
layer {
    bottom: "conv1_tnet2"
    top: "conv1_tnet2"
    name: "scale1_tnet2"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}
layer {
  name: "relu1_tnet2"
  type: "ReLU"
  bottom: "conv1_tnet2"
  top: "conv1_tnet2"
}

layer {
  name: "conv2_tnet2"
  type: "Convolution"
  bottom: "conv1_tnet2"
  top: "conv2_tnet2"
  param {
    lr_mult: 1
  }
  param {
    lr_mult: 2
  }
  convolution_param {
    num_output: 128
    pad: 0
    kernel_w: 1
    kernel_h: 1
    stride: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "bn2_tnet2"
  type: "BatchNorm"
  bottom: "conv2_tnet2"
  top: "conv2_tnet2"
}
layer {
    bottom: "conv2_tnet2"
    top: "conv2_tnet2"
    name: "scale2_tnet2"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}
layer {
  name: "relu2_tnet2"
  type: "ReLU"
  bottom: "conv2_tnet2"
  top: "conv2_tnet2"
}

layer {
  name: "conv3_tnet2"
  type: "Convolution"
  bottom: "conv2_tnet2"
  top: "conv3_tnet2"
  param {
    lr_mult: 1
  }
  param {
    lr_mult: 2
  }
  convolution_param {
    num_output: 1024
    pad: 0
    kernel_w: 1
    kernel_h: 1
    stride: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "bn3_tnet2"
  type: "BatchNorm"
  bottom: "conv3_tnet2"
  top: "conv3_tnet2"
}
layer {
    bottom: "conv3_tnet2"
    top: "conv3_tnet2"
    name: "scale3_tnet2"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}
layer {
  name: "relu3_tnet2"
  type: "ReLU"
  bottom: "conv3_tnet2"
  top: "conv3_tnet2"
}
layer {
  name: "pool_tnet2"
  type: "Pooling"
  bottom: "conv3_tnet2"
  top: "global_feat_tnet2"
  pooling_param {
    pool: MAX
    pad: 0
    kernel_h: 2048
    kernel_w: 1
    stride: 1
  }
}
layer {
  name: "fc1_tnet2"
  type: "InnerProduct"
  bottom: "global_feat_tnet2"
  top: "fc1_tnet2"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 1
    decay_mult: 1
  }
  inner_product_param {
    num_output: 512
    weight_filler {
      type: "gaussian"
      std: 0.001
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "bn6_tnet2"
  type: "BatchNorm"
  bottom: "fc1_tnet2"
  top: "fc1_tnet2"
}
layer {
    bottom: "fc1_tnet2"
    top: "fc1_tnet2"
    name: "scale6_tnet2"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}
layer {
  name: "relu6_tnet2"
  type: "ReLU"
  bottom: "fc1_tnet2"
  top: "fc1_tnet2"
}

layer {
  name: "fc2_tnet2"
  type: "InnerProduct"
  bottom: "fc1_tnet2"
  top: "fc2_tnet2"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 1
    decay_mult: 1
  }
  inner_product_param {
    num_output: 256
    weight_filler {
      type: "gaussian"
      std: 0.001
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "bn7_tnet2"
  type: "BatchNorm"
  bottom: "fc2_tnet2"
  top: "fc2_tnet2"
}
layer {
    bottom: "fc2_tnet2"
    top: "fc2_tnet2"
    name: "scale7_tnet2"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}
layer {
  name: "relu7_tnet2"
  type: "ReLU"
  bottom: "fc2_tnet2"
  top: "fc2_tnet2"
}
layer {
  name: "fc3_tnet2"
  type: "InnerProduct"
  bottom: "fc2_tnet2"
  top: "fc3_tnet2"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 1
    decay_mult: 1
  }
  inner_product_param {
    num_output: 4096 # 64*64
    weight_filler {
      type: "constant"
      value: 0
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
    name: "reshape_tnet2"
    type: "Reshape"
    bottom: "fc3_tnet2"
    top: "fc3_tnet2_reshape"
    reshape_param {
      shape {
		dim: -1
        dim: 64
        dim: 64
      }
    }
}
#layer {
#  name: "eye_tnet2"
#  type: "Python"
#  bottom: "fc3_tnet2"
#  top: "eye_tnet2"
#  python_param {
#    module: "eye_matrix_layer"
#    layer: "EyeMatrixLayer"
#	param_str: "{\'K\': 64}"
#  }
#}
layer {
  name: "eltwise_sum_tnet2"
  type: "Eltwise"
  bottom: "fc3_tnet2_reshape"
 # bottom: "eye_tnet2"
  top: "transform2"
  eltwise_param { operation: SUM }
}
# -------------------- feat transform ---------------------
layer {
  name: "perm_scale2"
  type: "Permute"
  bottom: "conv2" # Bx64xNx1
  top: "scale2_permute" # BxNx64x1
  permute_param {
    order: 0
    order: 2
    order: 1
    order: 3
  }
}
layer {
    name: "reshape_scale2"
    type: "Reshape"
    bottom: "scale2_permute"  # BxNx64x1
    top: "scale2_reshape"  # BxNx64
    reshape_param {
      shape {
        dim: 0  # copy the dimension from below
        dim: -1
        dim: 64 # infer it from the other dimensions
      }
    }
}
#layer {
#  name: "matmul_feat"
#  type: "MatrixMultiplication"
#  bottom: "scale2_reshape" # BxNx64
#  bottom: "transform2" # 64x64
#  top: "scale2_transform2" # BxNx64
#}
layer {
  name: "perm_scale2_transform2"
  type: "Permute"
  bottom: "scale2_transform2" # BxNx64
  top: "scale2_permute_transform2" # Bx64xN
  permute_param {
    order: 0
    order: 2
    order: 1
  }
}
layer {
    name: "reshape_scale2_transform2"
    type: "Reshape"
    bottom: "scale2_permute_transform2" # Bx64xN
    top: "scale2_transform2_reshape" # Bx64xNx1
    reshape_param {
      shape {
        dim: 0  # copy the dimension from below
        dim: 0
        dim: 0
        dim: 1 
      }
    }
}

# -------------------- mlp ---------------------
layer {
  name: "conv3"
  type: "Convolution"
  bottom: "scale2_transform2_reshape"
  top: "conv3"
  param {
    lr_mult: 1
  }
  param {
    lr_mult: 2
  }
  convolution_param {
    num_output: 64
    pad: 0
    kernel_w: 1
    kernel_h: 1
    stride: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "bn3"
  type: "BatchNorm"
  bottom: "conv3"
  top: "conv3"
}
layer {
    bottom: "conv3"
    top: "conv3"
    name: "scale3"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}
layer {
  name: "relu3"
  type: "ReLU"
  bottom: "conv3"
  top: "conv3"
}

layer {
  name: "conv4"
  type: "Convolution"
  bottom: "conv3"
  top: "conv4"
  param {
    lr_mult: 1
  }
  param {
    lr_mult: 2
  }
  convolution_param {
    num_output: 128
    pad: 0
    kernel_w: 1
    kernel_h: 1
    stride: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "bn4"
  type: "BatchNorm"
  bottom: "conv4"
  top: "conv4"
}
layer {
    bottom: "conv4"
    top: "conv4"
    name: "scale4"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}
layer {
  name: "relu4"
  type: "ReLU"
  bottom: "conv4"
  top: "conv4"
}

layer {
  name: "conv5"
  type: "Convolution"
  bottom: "conv4"
  top: "conv5"
  param {
    lr_mult: 1
  }
  param {
    lr_mult: 2
  }
  convolution_param {
    num_output: 1024
    pad: 0
    kernel_w: 1
    kernel_h: 1
    stride: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "bn5"
  type: "BatchNorm"
  bottom: "conv5"
  top: "conv5"
}
layer {
    bottom: "conv5"
    top: "conv5"
    name: "scale5"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}
layer {
  name: "relu5"
  type: "ReLU"
  bottom: "conv5"
  top: "conv5"
}

# -------------------- max pool ---------------------
layer {
  name: "pool"
  type: "Pooling"
  bottom: "conv5"
  top: "global_feat"
  pooling_param {
    pool: MAX
    pad: 0
    kernel_h: 2048
    kernel_w: 1
    stride: 1
  }
}

# -------------------- fc ---------------------
layer {
  name: "fc1"
  type: "InnerProduct"
  bottom: "global_feat"
  top: "fc1"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 1
    decay_mult: 1
  }
  inner_product_param {
    num_output: 512
    weight_filler {
      type: "gaussian"
      std: 0.001
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "bn6"
  type: "BatchNorm"
  bottom: "fc1"
  top: "fc1"
}
layer {
    bottom: "fc1"
    top: "fc1"
    name: "scale6"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}
layer {
  name: "relu6"
  type: "ReLU"
  bottom: "fc1"
  top: "fc1"
}

layer {
  name: "fc2"
  type: "InnerProduct"
  bottom: "fc1"
  top: "fc2"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 1
    decay_mult: 1
  }
  inner_product_param {
    num_output: 256
    weight_filler {
      type: "gaussian"
      std: 0.001
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
layer {
  name: "bn7"
  type: "BatchNorm"
  bottom: "fc2"
  top: "fc2"
}
layer {
    bottom: "fc2"
    top: "fc2"
    name: "scale7"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}
layer {
  name: "relu7"
  type: "ReLU"
  bottom: "fc2"
  top: "fc2"
}

layer {
  name: "drop1"
  type: "Dropout"
  bottom: "fc2"
  top: "drop1"
  dropout_param {
    dropout_ratio: 0.3
  }
}
layer {
  name: "fc3"
  type: "InnerProduct"
  bottom: "drop1"
  top: "fc3"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 1
    decay_mult: 1
  }
  inner_product_param {
    num_output: 40
    weight_filler {
      type: "gaussian"
      std: 0.001
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}
#layer {
 # name: "accuracy"
 # type: "Accuracy"
 # bottom: "fc3"
 # bottom: "label"
 # top: "accuracy"
 # include {
 #   phase: TEST
 # }
#}
layer {
  name: "loss"
  #type: "SoftmaxWithLoss"
  type:"Softmax"
  bottom: "fc3"
  bottom: "label"
  top: "loss"
}

# Enforce the transformation as orthogonal matrix
layer {
  name: "perm_loss_transform2"
  type: "Permute"
  bottom: "transform2" # Bx64x64
  top: "transform2_transpose" # Bx64x64
  permute_param {
    order: 0
    order: 2
    order: 1
  }
}
#layer {
 # name: "matmul_feat_loss"
 # type: "MatrixMultiplication"
 # bottom: "transform2"
 # bottom: "transform2_transpose"
 # top: "transform2_mul" # Bx64x64
#}
#layer {
#  name: "loss_feat"
#  type: "EuclideanLoss"
#  bottom: "transform2_mul"
#  bottom: "eye_tnet2"
#  top: "loss_feat"
#  loss_weight: 0.001
#}