from keras import models
from keras import layers
# Model

from keras.layers import LeakyReLU
network = models.Sequential()

network.add(layers.Conv2D(32, (3,3), activation=None, input_shape=(img_rows, img_cols, 1), padding='valid', use_bias=False,))
     network.add(LeakyReLU(alpha=0.1))

        network.add(layers.MaxPooling2D(pool_size=(2, 2)))

        network.add(layers.Conv2D(32, (3,3), activation=None,padding='valid', use_bias=True))

    network.add(LeakyReLU(alpha=0.1))

    network.add(layers.MaxPooling2D(pool_size=(2,2)))

    network.add(layers.Flatten())

    network.add(layers.Dense(64, activation='relu'))

    network.add(layers.Dense(3,  activation='softmax'))

    network.compile(optimizer='adam',
                      loss='categorical_crossentropy',
                      metrics=['accuracy'])

layer {
  name: "data"
  type: "Data"
  top: "data"
  top: "label"
  include {
    phase: TRAIN
  }
  data_param {
    source: "input/lmdb/train_lmdb"
    batch_size: 84
    backend: LMDB
  }
}
layer {
  name: "data"
  type: "Data"
  top: "data"
  top: "label"
  include {
    phase: TEST
  }
  data_param {
    source: "input/lmdb/valid_lmdb"
    batch_size: 84
    backend: LMDB
  }
}



layer {
  name: "conv1"
  type: "Convolution"
  bottom: "data"
  top: "conv1"
  convolution_param {
    num_output: 32
    pad: 0
    kernel_size: 3
    stride: 1
  }
}
layer {
  name: "lr1"
  type: "ReLU"
  bottom: "conv1"
  top: "lr1"
  relu_param {
    negative_slope: 0.10000000149
  }
}
layer {
  name: "mp1"
  type: "Pooling"
  bottom: "lr1"
  top: "mp1"
  pooling_param {
    pool: MAX
    kernel_size: 2
    stride: 1
    pad: 0
  }
}
layer {
  name: "conv2"
  type: "Convolution"
  bottom: "mp1"
  top: "conv2"
  convolution_param {
    num_output: 32
    pad: 0
    kernel_size: 3
    stride: 1
  }
}
layer {
  name: "lr2"
  type: "ReLU"
  bottom: "conv2"
  top: "lr2"
  relu_param {
    negative_slope: 0.10000000149
  }
}
layer {
  name: "mp2"
  type: "Pooling"
  bottom: "lr2"
  top: "mp2"
  pooling_param {
    pool: MAX
    kernel_size: 2
    stride: 1
    pad: 0
  }
}
layer {
  name: "fl1"
  type: "Flatten"
  bottom: "mp2"
  top: "fl1"
}


layer {
  name: "fc1"
  type: "InnerProduct"
  bottom: "fl1"
  top: "fc1"

  inner_product_param {
    num_output: 64
    weight_filler {
      type: "gaussian"
      #type: "xavier"
      std: 0.001
    }
    bias_filler {
      type: "constant"
      value: 1
    }
  }
}

layer {
  name: "relu1"
  type: "ReLU"
  bottom: "fc1"
  top: "relu1"

  }

layer {
  name: "fc2"
  type: "InnerProduct"
  bottom: "relu1"
  top: "fc2"

  inner_product_param {
    num_output: 3
    weight_filler {
      type: "gaussian"
      #type: "xavier"
      std: 0.001
    }
    bias_filler {
      type: "constant"
      value: 1
    }
  }
}
layer {
  name: "softmax"
  type: "Softmax"
  bottom: "fc2"
  top: "softmax"
}
layer {
  name: "accuracy"
  type: "Accuracy"
  bottom: "softmax"
  bottom: "label"
  top: "accuracy"
  include {
    phase: TEST
  }
}
layer {
  name: "loss"
  type: "MultinomialLogisticLoss"
  bottom: "softmax"
  bottom: "label"
  top: "loss"
}