CNN_M_2048_train_val.prototxt 文件

        最近在看的双流论文《Two-Stream Convolutional Networks for Action Recognition in Videos 》是基于CNN-M-2048 的网络结构的，本想在网上找了下相关的train_val.prototxt文件。最后只在ksimonyan/VGG_CNN_M_2048_deploy.prototxt里找到了一个deploy.prototxt文件和已经在 ILSVRC-2012 dataset上训练好的VGG_CNN_M_2048.caffemodel。根据这个deploy.prototxt文件结合以前的CaffeNet_train_val.prototxt文件。自己将CNN_M_2048_train_val.prototxt文件给补全如下 ：

name: "VGG_CNN_M_2048"
layer {
  name: "data"
  type: "Data"
  top: "data"
  top: "label"
  include {
    phase: TRAIN
  }
  transform_param {
    mirror: true
    crop_size: 224
    mean_file: "C:/WENTEST/caffe-master/examples/WEN/data/1210_1206/mean.binaryproto"
  }
  data_param {
    source: "C:/WENTEST/caffe-master/examples/WEN/data/1210_1206/train_lmdb"
    batch_size: 64
    backend: LMDB
  }
}
layer {
  name: "data"
  type: "Data"
  top: "data"
  top: "label"
  include {
    phase: TEST
  }
  transform_param {
    mirror: false
    crop_size: 224
    mean_file: "C:/WENTEST/caffe-master/examples/WEN/data/1210_1206/mean.binaryproto"
  }
  data_param {
    source: "C:/WENTEST/caffe-master/examples/WEN/data/1210_1206/test_lmdb"
    batch_size: 50
    backend: LMDB
  }
}
layer {
  bottom: "data"
  top: "conv1"
  name: "conv1"
  type: "Convolution"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
  convolution_param {
    num_output: 96
    kernel_size: 7
    stride: 2
  }
}
layer {
  bottom: "conv1"
  top: "conv1"
  name: "relu1"
  type: "ReLU"
}
layer {
  bottom: "conv1"
  top: "norm1"
  name: "norm1"
  type: "LRN"
  lrn_param {
    local_size: 5
    alpha: 0.0005
    beta: 0.75
    k: 2
  }
}
layer {
  bottom: "norm1"
  top: "pool1"
  name: "pool1"
  type: "Pooling"
  pooling_param {
    pool: MAX
    kernel_size: 3
    stride: 2
  }
}
layer {
  bottom: "pool1"
  top: "conv2"
  name: "conv2"
  type: "Convolution"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
  convolution_param {
    num_output: 256
    pad: 1
    kernel_size: 5
    stride: 2
  }
}
layer {
  bottom: "conv2"
  top: "conv2"
  name: "relu2"
  type: "ReLU"
}
layer {
  bottom: "conv2"
  top: "norm2"
  name: "norm2"
  type: "LRN"
  lrn_param {
    local_size: 5
    alpha: 0.0005
    beta: 0.75
    k: 2
  }
}
layer {
  bottom: "norm2"
  top: "pool2"
  name: "pool2"
  type: "Pooling"
  pooling_param {
    pool: MAX
    kernel_size: 3
    stride: 2
  }
}
layer {
  bottom: "pool2"
  top: "conv3"
  name: "conv3"
  type: "Convolution"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
  convolution_param {
    num_output: 512
    pad: 1
    kernel_size: 3
  }
}
layer {
  bottom: "conv3"
  top: "conv3"
  name: "relu3"
  type: "ReLU"
}
layer {
  bottom: "conv3"
  top: "conv4"
  name: "conv4"
  type: "Convolution"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
  convolution_param {
    num_output: 512
    pad: 1
    kernel_size: 3
  }
}
layer {
  bottom: "conv4"
  top: "conv4"
  name: "relu4"
  type: "ReLU"
}
layer {
  bottom: "conv4"
  top: "conv5"
  name: "conv5"
  type: "Convolution"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
  convolution_param {
    num_output: 512
    pad: 1
    kernel_size: 3
  }
}
layer {
  bottom: "conv5"
  top: "conv5"
  name: "relu5"
  type: "ReLU"
}
layer {
  bottom: "conv5"
  top: "pool5"
  name: "pool5"
  type: "Pooling"
  pooling_param {
    pool: MAX
    kernel_size: 3
    stride: 2
  }
}
layer {
  bottom: "pool5"
  top: "fc6"
  name: "fc6"
  type: "InnerProduct"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
  inner_product_param {
    num_output: 4096
  }
}
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: "InnerProduct"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  param {
    lr_mult: 0
    decay_mult: 0
  }
  inner_product_param {
    num_output: 2048
  }
}
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_w"
  name: "fc8_w"
  type: "InnerProduct"
  param {
    lr_mult: 10
    decay_mult: 1
  }
  param {
    lr_mult: 20
    decay_mult: 0
  }
  inner_product_param {
    num_output: 51
  }
}
layer {
  name: "accuracy"
  type: "Accuracy"
  bottom: "fc8_w"
  bottom: "label"
  top: "accuracy"
  include {
    phase: TEST
  }
}
layer {
  name: "loss"
  type: "SoftmaxWithLoss"
  bottom: "fc8_w"
  bottom: "label"
  top: "loss"
} 

利用训练好的caffemodel在最后一层进行微调。固定前面层（学习率被我置为0），修改最后一层名称（fc8改为fc8_w）和输出类别数（1000改为51）。