MNTDP base architecture.PROTOTXT

# In order to easily visualize, and to obtain the number of parameters and the number of operations required for one forward pass, for the base architecture used by MNTDP, insert this Caffe Prototxt code into: http://dgschwend.github.io/netscope/#/editor (Accessed on 4 May 2022)
# Note that the definition of the network below is based on the input of one sample RGB image of size 32x32 and on the output of one vector of size 10, i.e. the number of classes for each considered dataset, as per MNTDP
# Furthermore, note that in the network definition below indications of the individual (trainable) modules are given, including indications of their constituting blocks of layers, for clarity



name: "MNTDP base architecture"

# =============================
# === Input - Start
# =============================
layer {
    name: "input"
    type: "Input"
    top: "input"
    input_param { shape: { dim: 1 dim: 3 dim: 32 dim: 32 } }
}
# =============================
# === Input - End
# =============================



# =============================
# === Module 1 - Start
# =============================
layer {
    name: "conv1"
    type: "Convolution"
    bottom: "input"
    top: "conv1"
    convolution_param {
        num_output: 64
        kernel_size: 3
        stride: 1
        pad: 1
        bias_term: false
    }
}
layer {
    name: "bn_conv1"
    type: "BatchNorm"
    bottom: "conv1"
    top: "conv1"
    batch_norm_param {
        moving_average_fraction: 0.9
        eps: 0.00001
    }
}
layer {
    name: "scale_conv1"
    type: "Scale"
    bottom: "conv1"
    top: "conv1"
    scale_param {
        bias_term: true
    }
}
layer {
    name: "relu_conv1"
    type: "ReLU"
    bottom: "conv1"
    top: "conv1"
}
# =============================
# === Module 1 - End
# =============================



# =============================
# === Module 2 - Start
# =============================
# --- Module 2, Block 1 - Start
layer {
    name: "conv2_1_1"
    type: "Convolution"
    bottom: "conv1"
    top: "conv2_1_1"
    convolution_param {
        num_output: 64
        kernel_size: 3
        stride: 1
        pad: 1
        bias_term: false
    }
}
layer {
    name: "bn_conv2_1_1"
    type: "BatchNorm"
    bottom: "conv2_1_1"
    top: "conv2_1_1"
    batch_norm_param {
        moving_average_fraction: 0.9
        eps: 0.00001
    }
}
layer {
    name: "scale_conv2_1_1"
    type: "Scale"
    bottom: "conv2_1_1"
    top: "conv2_1_1"
    scale_param {
        bias_term: true
    }
}
layer {
    name: "relu_conv2_1_1"
    type: "ReLU"
    bottom: "conv2_1_1"
    top: "conv2_1_1"
}
layer {
    name: "conv2_1_2"
    type: "Convolution"
    bottom: "conv2_1_1"
    top: "conv2_1_1"
    convolution_param {
        num_output: 64
        kernel_size: 3
        stride: 1
        pad: 1
        bias_term: false
    }
}
layer {
    name: "bn_conv2_1_2"
    type: "BatchNorm"
    bottom: "conv2_1_1"
    top: "conv2_1_1"
    batch_norm_param {
        moving_average_fraction: 0.9
        eps: 0.00001
    }
}
layer {
    name: "scale_conv2_1_2"
    type: "Scale"
    bottom: "conv2_1_1"
    top: "conv2_1_1"
    scale_param {
        bias_term: true
    }
}
# --- Module 2, Block 1 - End
# --- Module 2, Block 2 - Start
layer {
    name: "conv2_2_1"
    type: "Convolution"
    bottom: "conv2_1_1"
    top: "conv2_2_1"
    convolution_param {
        num_output: 64
        kernel_size: 3
        stride: 1
        pad: 1
        bias_term: false
    }
}
layer {
    name: "bn_conv2_2_1"
    type: "BatchNorm"
    bottom: "conv2_2_1"
    top: "conv2_2_1"
    batch_norm_param {
        moving_average_fraction: 0.9
        eps: 0.00001
    }
}
layer {
    name: "scale_conv2_2_1"
    type: "Scale"
    bottom: "conv2_2_1"
    top: "conv2_2_1"
    scale_param {
        bias_term: true
    }
}
layer {
    name: "relu_conv2_2_1"
    type: "ReLU"
    bottom: "conv2_2_1"
    top: "conv2_2_1"
}
layer {
    name: "conv2_2_2"
    type: "Convolution"
    bottom: "conv2_2_1"
    top: "conv2_2_1"
    convolution_param {
        num_output: 64
        kernel_size: 3
        stride: 1
        pad: 1
        bias_term: false
    }
}
layer {
    name: "bn_conv2_2_2"
    type: "BatchNorm"
    bottom: "conv2_2_1"
    top: "conv2_2_1"
    batch_norm_param {
        moving_average_fraction: 0.9
        eps: 0.00001
    }
}
layer {
    name: "scale_conv2_2_2"
    type: "Scale"
    bottom: "conv2_2_1"
    top: "conv2_2_1"
    scale_param {
        bias_term: true
    }
}
# --- Module 2, Block 2 - End
# =============================
# === Module 2 - End
# =============================



# =============================
# === Module 3 - Start
# =============================
# --- Module 3, Block 1 - Start
layer {
    name: "conv3_1_1"
    type: "Convolution"
    bottom: "conv2_2_1"
    top: "conv3_1_1"
    convolution_param {
        num_output: 64
        kernel_size: 3
        stride: 2
        pad: 1
        bias_term: false
    }
}
layer {
    name: "bn_conv3_1_1"
    type: "BatchNorm"
    bottom: "conv3_1_1"
    top: "conv3_1_1"
    batch_norm_param {
        moving_average_fraction: 0.9
        eps: 0.00001
    }
}
layer {
    name: "scale_conv3_1_1"
    type: "Scale"
    bottom: "conv3_1_1"
    top: "conv3_1_1"
    scale_param {
        bias_term: true
    }
}
layer {
    name: "relu_conv3_1_1"
    type: "ReLU"
    bottom: "conv3_1_1"
    top: "conv3_1_1"
}
layer {
    name: "conv3_1_2"
    type: "Convolution"
    bottom: "conv3_1_1"
    top: "conv3_1_1"
    convolution_param {
        num_output: 64
        kernel_size: 3
        stride: 1
        pad: 1
        bias_term: false
    }
}
layer {
    name: "bn_conv3_1_2"
    type: "BatchNorm"
    bottom: "conv3_1_1"
    top: "conv3_1_1"
    batch_norm_param {
        moving_average_fraction: 0.9
        eps: 0.00001
    }
}
layer {
    name: "scale_conv3_1_2"
    type: "Scale"
    bottom: "conv3_1_1"
    top: "conv3_1_1"
    scale_param {
        bias_term: true
    }
}
# --- Module 3, Block 1, Downsample - Start
layer {
    name: "conv3_d"
    type: "Convolution"
    bottom: "conv2_2_1"
    top: "conv3_d"
    convolution_param {
        num_output: 64
        kernel_size: 1
        stride: 2
        bias_term: false
    }
}
layer {
    name: "bn_conv3_d"
    type: "BatchNorm"
    bottom: "conv3_d"
    top: "conv3_d"
    batch_norm_param {
        moving_average_fraction: 0.9
        eps: 0.00001
    }
}
layer {
    name: "scale_conv3_d"
    type: "Scale"
    bottom: "conv3_d"
    top: "conv3_d"
    scale_param {
        bias_term: true
    }
}
# --- Module 3, Block 1, Downsample - End
layer {
    bottom: "conv3_1_1"
    bottom: "conv3_d"
    top: "eltwise3"
    name: "eltwise3"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}
# --- Module 3, Block 1 - End
# --- Module 3, Block 2 - Start
layer {
    name: "conv3_2_1"
    type: "Convolution"
    bottom: "eltwise3"
    top: "conv3_2_1"
    convolution_param {
        num_output: 64
        kernel_size: 3
        stride: 1
        pad: 1
        bias_term: false
    }
}
layer {
    name: "bn_conv3_2_1"
    type: "BatchNorm"
    bottom: "conv3_2_1"
    top: "conv3_2_1"
    batch_norm_param {
        moving_average_fraction: 0.9
        eps: 0.00001
    }
}
layer {
    name: "scale_conv3_2_1"
    type: "Scale"
    bottom: "conv3_2_1"
    top: "conv3_2_1"
    scale_param {
        bias_term: true
    }
}
layer {
    name: "relu_conv3_2_1"
    type: "ReLU"
    bottom: "conv3_2_1"
    top: "conv3_2_1"
}
layer {
    name: "conv3_2_2"
    type: "Convolution"
    bottom: "conv3_2_1"
    top: "conv3_2_1"
    convolution_param {
        num_output: 64
        kernel_size: 3
        stride: 1
        pad: 1
        bias_term: false
    }
}
layer {
    name: "bn_conv3_2_2"
    type: "BatchNorm"
    bottom: "conv3_2_1"
    top: "conv3_2_1"
    batch_norm_param {
        moving_average_fraction: 0.9
        eps: 0.00001
    }
}
layer {
    name: "scale_conv3_2_2"
    type: "Scale"
    bottom: "conv3_2_1"
    top: "conv3_2_1"
    scale_param {
        bias_term: true
    }
}
# --- Module 3, Block 2 - End
# =============================
# === Module 3 - End
# =============================



# =============================
# === Module 4 - Start
# =============================
# --- Module 4, Block 1 - Start
layer {
    name: "conv4_1_1"
    type: "Convolution"
    bottom: "conv3_2_1"
    top: "conv4_1_1"
    convolution_param {
        num_output: 64
        kernel_size: 3
        stride: 2
        pad: 1
        bias_term: false
    }
}
layer {
    name: "bn_conv4_1_1"
    type: "BatchNorm"
    bottom: "conv4_1_1"
    top: "conv4_1_1"
    batch_norm_param {
        moving_average_fraction: 0.9
        eps: 0.00001
    }
}
layer {
    name: "scale_conv4_1_1"
    type: "Scale"
    bottom: "conv4_1_1"
    top: "conv4_1_1"
    scale_param {
        bias_term: true
    }
}
layer {
    name: "relu_conv4_1_1"
    type: "ReLU"
    bottom: "conv4_1_1"
    top: "conv4_1_1"
}
layer {
    name: "conv4_1_2"
    type: "Convolution"
    bottom: "conv4_1_1"
    top: "conv4_1_1"
    convolution_param {
        num_output: 64
        kernel_size: 3
        stride: 1
        pad: 1
        bias_term: false
    }
}
layer {
    name: "bn_conv4_1_2"
    type: "BatchNorm"
    bottom: "conv4_1_1"
    top: "conv4_1_1"
    batch_norm_param {
        moving_average_fraction: 0.9
        eps: 0.00001
    }
}
layer {
    name: "scale_conv4_1_2"
    type: "Scale"
    bottom: "conv4_1_1"
    top: "conv4_1_1"
    scale_param {
        bias_term: true
    }
}
# --- Module 4, Block 1, Downsample - Start
layer {
    name: "conv4_d"
    type: "Convolution"
    bottom: "conv3_2_1"
    top: "conv4_d"
    convolution_param {
        num_output: 64
        kernel_size: 1
        stride: 2
        bias_term: false
    }
}
layer {
    name: "bn_conv4_d"
    type: "BatchNorm"
    bottom: "conv4_d"
    top: "conv4_d"
    batch_norm_param {
        moving_average_fraction: 0.9
        eps: 0.00001
    }
}
layer {
    name: "scale_conv4_d"
    type: "Scale"
    bottom: "conv4_d"
    top: "conv4_d"
    scale_param {
        bias_term: true
    }
}
# --- Module 4, Block 1, Downsample - End
layer {
    bottom: "conv4_1_1"
    bottom: "conv4_d"
    top: "eltwise4"
    name: "eltwise4"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}
# --- Module 4, Block 1 - End
# --- Module 4, Block 2 - Start
layer {
    name: "conv4_2_1"
    type: "Convolution"
    bottom: "eltwise4"
    top: "conv4_2_1"
    convolution_param {
        num_output: 64
        kernel_size: 3
        stride: 1
        pad: 1
        bias_term: false
    }
}
layer {
    name: "bn_conv4_2_1"
    type: "BatchNorm"
    bottom: "conv4_2_1"
    top: "conv4_2_1"
    batch_norm_param {
        moving_average_fraction: 0.9
        eps: 0.00001
    }
}
layer {
    name: "scale_conv4_2_1"
    type: "Scale"
    bottom: "conv4_2_1"
    top: "conv4_2_1"
    scale_param {
        bias_term: true
    }
}
layer {
    name: "relu_conv4_2_1"
    type: "ReLU"
    bottom: "conv4_2_1"
    top: "conv4_2_1"
}
layer {
    name: "conv4_2_2"
    type: "Convolution"
    bottom: "conv4_2_1"
    top: "conv4_2_1"
    convolution_param {
        num_output: 64
        kernel_size: 3
        stride: 1
        pad: 1
        bias_term: false
    }
}
layer {
    name: "bn_conv4_2_2"
    type: "BatchNorm"
    bottom: "conv4_2_1"
    top: "conv4_2_1"
    batch_norm_param {
        moving_average_fraction: 0.9
        eps: 0.00001
    }
}
layer {
    name: "scale_conv4_2_2"
    type: "Scale"
    bottom: "conv4_2_1"
    top: "conv4_2_1"
    scale_param {
        bias_term: true
    }
}
# --- Module 4, Block 2 - End
# =============================
# === Module 4 - End
# =============================



# =============================
# === Module 5 - Start
# =============================
layer {
    name: "conv5_1"
    type: "Convolution"
    bottom: "conv4_2_1"
    top: "conv5_1"
    convolution_param {
        num_output: 64
        kernel_size: 3
        stride: 2
        pad: 1
        bias_term: false
    }
}
layer {
    name: "bn_conv5_1"
    type: "BatchNorm"
    bottom: "conv5_1"
    top: "conv5_1"
    batch_norm_param {
        moving_average_fraction: 0.9
        eps: 0.00001
    }
}
layer {
    name: "scale_conv5_1"
    type: "Scale"
    bottom: "conv5_1"
    top: "conv5_1"
    scale_param {
        bias_term: true
    }
}
layer {
    name: "relu_conv5_1"
    type: "ReLU"
    bottom: "conv5_1"
    top: "conv5_1"
}
layer {
    name: "conv5_2"
    type: "Convolution"
    bottom: "conv5_1"
    top: "conv5_1"
    convolution_param {
        num_output: 64
        kernel_size: 3
        stride: 1
        pad: 1
        bias_term: false
    }
}
layer {
    name: "bn_conv5_2"
    type: "BatchNorm"
    bottom: "conv5_1"
    top: "conv5_1"
    batch_norm_param {
        moving_average_fraction: 0.9
        eps: 0.00001
    }
}
layer {
    name: "scale_conv5_2"
    type: "Scale"
    bottom: "conv5_1"
    top: "conv5_1"
    scale_param {
        bias_term: true
    }
}
# --- Module 5, Downsample - Start
layer {
    name: "conv5_d"
    type: "Convolution"
    bottom: "conv4_2_1"
    top: "conv5_d"
    convolution_param {
        num_output: 64
        kernel_size: 1
        stride: 2
        bias_term: false
    }
}
layer {
    name: "bn_conv5_d"
    type: "BatchNorm"
    bottom: "conv5_d"
    top: "conv5_d"
    batch_norm_param {
        moving_average_fraction: 0.9
        eps: 0.00001
    }
}
layer {
    name: "scale_conv5_d"
    type: "Scale"
    bottom: "conv5_d"
    top: "conv5_d"
    scale_param {
        bias_term: true
    }
}
# --- Module 5, Downsample - End
layer {
    bottom: "conv5_1"
    bottom: "conv5_d"
    top: "eltwise5"
    name: "eltwise5"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}
# =============================
# === Module 5 - End
# =============================



# =============================
# === Module 6 - Start
# =============================
layer {
    name: "conv6_1"
    type: "Convolution"
    bottom: "eltwise5"
    top: "conv6_1"
    convolution_param {
        num_output: 64
        kernel_size: 3
        stride: 1
        pad: 1
        bias_term: false
    }
}
layer {
    name: "bn_conv6_1"
    type: "BatchNorm"
    bottom: "conv6_1"
    top: "conv6_1"
    batch_norm_param {
        moving_average_fraction: 0.9
        eps: 0.00001
    }
}
layer {
    name: "scale_conv6_1"
    type: "Scale"
    bottom: "conv6_1"
    top: "conv6_1"
    scale_param {
        bias_term: true
    }
}
layer {
    name: "relu_conv6_1"
    type: "ReLU"
    bottom: "conv6_1"
    top: "conv6_1"
}
layer {
    name: "conv6_2"
    type: "Convolution"
    bottom: "conv6_1"
    top: "conv6_1"
    convolution_param {
        num_output: 64
        kernel_size: 3
        stride: 1
        pad: 1
        bias_term: false
    }
}
layer {
    name: "bn_conv6_2"
    type: "BatchNorm"
    bottom: "conv6_1"
    top: "conv6_1"
    batch_norm_param {
        moving_average_fraction: 0.9
        eps: 0.00001
    }
}
layer {
    name: "scale_conv6_2"
    type: "Scale"
    bottom: "conv6_1"
    top: "conv6_1"
    scale_param {
        bias_term: true
    }
}
# --- Module 6, Block 1, Pooling - Start
layer {
    name: "relu6_2_a"
    type: "ReLU"
    bottom: "conv6_1"
    top: "relu6_2_a"
}
layer {
    name: "pool6_2_a"
    type: "Pooling"
    bottom: "relu6_2_a"
    top: "relu6_2_a"
    pooling_param {
        kernel_size: 4
        stride: 1
        pool: AVE
    }
}
# --- Module 6, Block 1, Pooling - End
# =============================
# === Module 6 - End
# =============================



# =============================
# === Module 7 - Start
# =============================
layer {
    name: "flatten"
    type: "Flatten"
    bottom: "relu6_2_a"
    top: "outputlayer"
}
layer {
    name: "linear"
    type: "InnerProduct"
    bottom: "outputlayer"
    top: "outputlayer"
    inner_product_param {
      	num_output: 10
    }
}
# =============================
# === Module 7 - End
# =============================