[딥 러 닝]노트 5caffe 데이터 계층 파라미터 에 대한 상세 한 설명

/**********************************************************************************************
*    :
*       [      5]
*              caffe        
*    :
*      UBUNTU14.04+caffe+python     
*    :
*              ----2016.10.31
*    :
*        http://caffe.berkeleyvision.org/gathered/examples/cifar10.html
**********************************************************************************************/
( )caffe    :    
    caffe      Blobs(  caffe       )    .[   ] [  ]   Blob    , blob
   [    ]    .
        (     ),    (    [0,255]         [0,1])         [   ]
    .
                [   ]。
( )      
layer {                             
  name: "mnist"            //[1]       --mnist
  type: "Data"             //[2]     levelDB LMDB              
  top: "data"              //[3]      blob   :data,label(     CNN  )
  top: "label"                      
  include {                //[4]include       ,          
    phase: TRAIN
  }
  transform_param {        //[5]     ,       
    scale: 0.00390625         //[1]       ,    [0,255] MNIST      [0,1]
                              //[2]       
    mean_file: "examples/myself/myself_mean.binaryproto"
    mirror:1                  //[3] images                ,   :         
                              //   #1=on,0=off
    crop_size:227             //[4]       crop_size*crop_size
  }
  data_param {            //[6]      
    source: "examples/mnist/mnist_train_lmdb"
                               //[1]     
    batch_size: 64             //[2]    ,  caffe      LMDB        
    backend: LMDB              //[3]           LevelDB   
  }
}