Caffe代码导读(5):对数据集进行Testing

上一篇介绍了如何准备数据集,做好准备之后我们先看怎样对训练好的模型进行Testing。


先用手写体识别例子,MNIST是数据集(包括训练数据和测试数据),深度学习模型采用LeNet(具体介绍见http://yann.lecun.com/exdb/lenet/),由Yann LeCun教授提出。


如果你编译好了Caffe,那么在CAFFE_ROOT下运行如下命令:


$ ./build/tools/caffe.bin test -model=examples/mnist/lenet_train_test.prototxt -weights=examples/mnist/lenet_iter_10000.caffemodel -gpu=0

就可以实现Testing。参数说明如下:

test:表示对训练好的模型进行Testing,而不是training。其他参数包括train, time, device_query。

-model=XXX:指定模型prototxt文件,这是一个文本文件,详细描述了网络结构和数据集信息。我用的prototxt内容如下:

name: "LeNet"
layers {
  name: "mnist"
  type: DATA
  top: "data"
  top: "label"
  data_param {
    source: "examples/mnist/mnist_train_lmdb"
    backend: LMDB
    batch_size: 64
  }
  transform_param {
    scale: 0.00390625
  }
  include: { phase: TRAIN }
}
layers {
  name: "mnist"
  type: DATA
  top: "data"
  top: "label"
  data_param {
    source: "examples/mnist/mnist_test_lmdb"
    backend: LMDB
    batch_size: 100
  }
  transform_param {
    scale: 0.00390625
  }
  include: { phase: TEST }
}

layers {
  name: "conv1"
  type: CONVOLUTION
  bottom: "data"
  top: "conv1"
  blobs_lr: 1
  blobs_lr: 2
  convolution_param {
    num_output: 20
    kernel_size: 5
    stride: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
    }
  }
}
layers {
  name: "pool1"
  type: POOLING
  bottom: "conv1"
  top: "pool1"
  pooling_param {
    pool: MAX
    kernel_size: 2
    stride: 2
  }
}
layers {
  name: "conv2"
  type: CONVOLUTION
  bottom: "pool1"
  top: "conv2"
  blobs_lr: 1
  blobs_lr: 2
  convolution_param {
    num_output: 50
    kernel_size: 5
    stride: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
    }
  }
}
layers {
  name: "pool2"
  type: POOLING
  bottom: "conv2"
  top: "pool2"
  pooling_param {
    pool: MAX
    kernel_size: 2
    stride: 2
  }
}
layers {
  name: "ip1"
  type: INNER_PRODUCT
  bottom: "pool2"
  top: "ip1"
  blobs_lr: 1
  blobs_lr: 2
  inner_product_param {
    num_output: 500
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
    }
  }
}
layers {
  name: "relu1"
  type: RELU
  bottom: "ip1"
  top: "ip1"
}
layers {
  name: "ip2"
  type: INNER_PRODUCT
  bottom: "ip1"
  top: "ip2"
  blobs_lr: 1
  blobs_lr: 2
  inner_product_param {
    num_output: 10
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
    }
  }
}
layers {
  name: "accuracy"
  type: ACCURACY
  bottom: "ip2"
  bottom: "label"
  top: "accuracy"
  include: { phase: TEST }
}
layers {
  name: "loss"
  type: SOFTMAX_LOSS
  bottom: "ip2"
  bottom: "label"
  top: "loss"
}

里面定义的网络结构如下图所示:

Caffe代码导读(5):对数据集进行Testing_第1张图片


-weights=XXX:指定训练好的caffemodel二进制文件。如果你手头没有训练好的可以下载这个(http://download.csdn.net/detail/kkk584520/8219443)。

-gpu=0:指定在GPU上运行,GPUID=0。如果你没有GPU就去掉这个参数,默认在CPU上运行。


运行输出如下:

I1203 18:47:00.073052  4610 caffe.cpp:134] Use GPU with device ID 0
I1203 18:47:00.367065  4610 net.cpp:275] The NetState phase (1) differed from the phase (0) specified by a rule in layer mnist
I1203 18:47:00.367269  4610 net.cpp:39] Initializing net from parameters: 
name: "LeNet"
layers {
  top: "data"
  top: "label"
  name: "mnist"
  type: DATA
  data_param {
    source: "examples/mnist/mnist_test_lmdb"
    batch_size: 100
    backend: LMDB
  }
  include {
    phase: TEST
  }
  transform_param {
    scale: 0.00390625
  }
}
layers {
  bottom: "data"
  top: "conv1"
  name: "conv1"
  type: CONVOLUTION
  blobs_lr: 1
  blobs_lr: 2
  convolution_param {
    num_output: 20
    kernel_size: 5
    stride: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
    }
  }
}
layers {
  bottom: "conv1"
  top: "pool1"
  name: "pool1"
  type: POOLING
  pooling_param {
    pool: MAX
    kernel_size: 2
    stride: 2
  }
}
layers {
  bottom: "pool1"
  top: "conv2"
  name: "conv2"
  type: CONVOLUTION
  blobs_lr: 1
  blobs_lr: 2
  convolution_param {
    num_output: 50
    kernel_size: 5
    stride: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
    }
  }
}
layers {
  bottom: "conv2"
  top: "pool2"
  name: "pool2"
  type: POOLING
  pooling_param {
    pool: MAX
    kernel_size: 2
    stride: 2
  }
}
layers {
  bottom: "pool2"
  top: "ip1"
  name: "ip1"
  type: INNER_PRODUCT
  blobs_lr: 1
  blobs_lr: 2
  inner_product_param {
    num_output: 500
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
    }
  }
}
layers {
  bottom: "ip1"
  top: "ip1"
  name: "relu1"
  type: RELU
}
layers {
  bottom: "ip1"
  top: "ip2"
  name: "ip2"
  type: INNER_PRODUCT
  blobs_lr: 1
  blobs_lr: 2
  inner_product_param {
    num_output: 10
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
    }
  }
}
layers {
  bottom: "ip2"
  bottom: "label"
  top: "accuracy"
  name: "accuracy"
  type: ACCURACY
  include {
    phase: TEST
  }
}
layers {
  bottom: "ip2"
  bottom: "label"
  top: "loss"
  name: "loss"
  type: SOFTMAX_LOSS
}
I1203 18:47:00.367391  4610 net.cpp:67] Creating Layer mnist
I1203 18:47:00.367409  4610 net.cpp:356] mnist -> data
I1203 18:47:00.367435  4610 net.cpp:356] mnist -> label
I1203 18:47:00.367451  4610 net.cpp:96] Setting up mnist
I1203 18:47:00.367571  4610 data_layer.cpp:68] Opening lmdb examples/mnist/mnist_test_lmdb
I1203 18:47:00.367609  4610 data_layer.cpp:128] output data size: 100,1,28,28
I1203 18:47:00.367832  4610 net.cpp:103] Top shape: 100 1 28 28 (78400)
I1203 18:47:00.367849  4610 net.cpp:103] Top shape: 100 1 1 1 (100)
I1203 18:47:00.367863  4610 net.cpp:67] Creating Layer label_mnist_1_split
I1203 18:47:00.367873  4610 net.cpp:394] label_mnist_1_split <- label
I1203 18:47:00.367892  4610 net.cpp:356] label_mnist_1_split -> label_mnist_1_split_0
I1203 18:47:00.367908  4610 net.cpp:356] label_mnist_1_split -> label_mnist_1_split_1
I1203 18:47:00.367919  4610 net.cpp:96] Setting up label_mnist_1_split
I1203 18:47:00.367929  4610 net.cpp:103] Top shape: 100 1 1 1 (100)
I1203 18:47:00.367938  4610 net.cpp:103] Top shape: 100 1 1 1 (100)
I1203 18:47:00.367950  4610 net.cpp:67] Creating Layer conv1
I1203 18:47:00.367959  4610 net.cpp:394] conv1 <- data
I1203 18:47:00.367969  4610 net.cpp:356] conv1 -> conv1
I1203 18:47:00.367982  4610 net.cpp:96] Setting up conv1
I1203 18:47:00.392133  4610 net.cpp:103] Top shape: 100 20 24 24 (1152000)
I1203 18:47:00.392204  4610 net.cpp:67] Creating Layer pool1
I1203 18:47:00.392217  4610 net.cpp:394] pool1 <- conv1
I1203 18:47:00.392231  4610 net.cpp:356] pool1 -> pool1
I1203 18:47:00.392247  4610 net.cpp:96] Setting up pool1
I1203 18:47:00.392273  4610 net.cpp:103] Top shape: 100 20 12 12 (288000)
I1203 18:47:00.392297  4610 net.cpp:67] Creating Layer conv2
I1203 18:47:00.392307  4610 net.cpp:394] conv2 <- pool1
I1203 18:47:00.392318  4610 net.cpp:356] conv2 -> conv2
I1203 18:47:00.392330  4610 net.cpp:96] Setting up conv2
I1203 18:47:00.392669  4610 net.cpp:103] Top shape: 100 50 8 8 (320000)
I1203 18:47:00.392729  4610 net.cpp:67] Creating Layer pool2
I1203 18:47:00.392756  4610 net.cpp:394] pool2 <- conv2
I1203 18:47:00.392768  4610 net.cpp:356] pool2 -> pool2
I1203 18:47:00.392781  4610 net.cpp:96] Setting up pool2
I1203 18:47:00.392793  4610 net.cpp:103] Top shape: 100 50 4 4 (80000)
I1203 18:47:00.392810  4610 net.cpp:67] Creating Layer ip1
I1203 18:47:00.392819  4610 net.cpp:394] ip1 <- pool2
I1203 18:47:00.392832  4610 net.cpp:356] ip1 -> ip1
I1203 18:47:00.392844  4610 net.cpp:96] Setting up ip1
I1203 18:47:00.397348  4610 net.cpp:103] Top shape: 100 500 1 1 (50000)
I1203 18:47:00.397372  4610 net.cpp:67] Creating Layer relu1
I1203 18:47:00.397382  4610 net.cpp:394] relu1 <- ip1
I1203 18:47:00.397394  4610 net.cpp:345] relu1 -> ip1 (in-place)
I1203 18:47:00.397407  4610 net.cpp:96] Setting up relu1
I1203 18:47:00.397420  4610 net.cpp:103] Top shape: 100 500 1 1 (50000)
I1203 18:47:00.397434  4610 net.cpp:67] Creating Layer ip2
I1203 18:47:00.397442  4610 net.cpp:394] ip2 <- ip1
I1203 18:47:00.397456  4610 net.cpp:356] ip2 -> ip2
I1203 18:47:00.397469  4610 net.cpp:96] Setting up ip2
I1203 18:47:00.397532  4610 net.cpp:103] Top shape: 100 10 1 1 (1000)
I1203 18:47:00.397547  4610 net.cpp:67] Creating Layer ip2_ip2_0_split
I1203 18:47:00.397557  4610 net.cpp:394] ip2_ip2_0_split <- ip2
I1203 18:47:00.397565  4610 net.cpp:356] ip2_ip2_0_split -> ip2_ip2_0_split_0
I1203 18:47:00.397583  4610 net.cpp:356] ip2_ip2_0_split -> ip2_ip2_0_split_1
I1203 18:47:00.397593  4610 net.cpp:96] Setting up ip2_ip2_0_split
I1203 18:47:00.397603  4610 net.cpp:103] Top shape: 100 10 1 1 (1000)
I1203 18:47:00.397611  4610 net.cpp:103] Top shape: 100 10 1 1 (1000)
I1203 18:47:00.397622  4610 net.cpp:67] Creating Layer accuracy
I1203 18:47:00.397631  4610 net.cpp:394] accuracy <- ip2_ip2_0_split_0
I1203 18:47:00.397640  4610 net.cpp:394] accuracy <- label_mnist_1_split_0
I1203 18:47:00.397650  4610 net.cpp:356] accuracy -> accuracy
I1203 18:47:00.397661  4610 net.cpp:96] Setting up accuracy
I1203 18:47:00.397673  4610 net.cpp:103] Top shape: 1 1 1 1 (1)
I1203 18:47:00.397687  4610 net.cpp:67] Creating Layer loss
I1203 18:47:00.397696  4610 net.cpp:394] loss <- ip2_ip2_0_split_1
I1203 18:47:00.397706  4610 net.cpp:394] loss <- label_mnist_1_split_1
I1203 18:47:00.397714  4610 net.cpp:356] loss -> loss
I1203 18:47:00.397725  4610 net.cpp:96] Setting up loss
I1203 18:47:00.397737  4610 net.cpp:103] Top shape: 1 1 1 1 (1)
I1203 18:47:00.397745  4610 net.cpp:109]     with loss weight 1
I1203 18:47:00.397776  4610 net.cpp:170] loss needs backward computation.
I1203 18:47:00.397785  4610 net.cpp:172] accuracy does not need backward computation.
I1203 18:47:00.397794  4610 net.cpp:170] ip2_ip2_0_split needs backward computation.
I1203 18:47:00.397801  4610 net.cpp:170] ip2 needs backward computation.
I1203 18:47:00.397809  4610 net.cpp:170] relu1 needs backward computation.
I1203 18:47:00.397816  4610 net.cpp:170] ip1 needs backward computation.
I1203 18:47:00.397825  4610 net.cpp:170] pool2 needs backward computation.
I1203 18:47:00.397832  4610 net.cpp:170] conv2 needs backward computation.
I1203 18:47:00.397843  4610 net.cpp:170] pool1 needs backward computation.
I1203 18:47:00.397851  4610 net.cpp:170] conv1 needs backward computation.
I1203 18:47:00.397860  4610 net.cpp:172] label_mnist_1_split does not need backward computation.
I1203 18:47:00.397867  4610 net.cpp:172] mnist does not need backward computation.
I1203 18:47:00.397874  4610 net.cpp:208] This network produces output accuracy
I1203 18:47:00.397884  4610 net.cpp:208] This network produces output loss
I1203 18:47:00.397905  4610 net.cpp:467] Collecting Learning Rate and Weight Decay.
I1203 18:47:00.397915  4610 net.cpp:219] Network initialization done.
I1203 18:47:00.397923  4610 net.cpp:220] Memory required for data: 8086808
I1203 18:47:00.432165  4610 caffe.cpp:145] Running for 50 iterations.
I1203 18:47:00.435849  4610 caffe.cpp:169] Batch 0, accuracy = 0.99
I1203 18:47:00.435879  4610 caffe.cpp:169] Batch 0, loss = 0.018971
I1203 18:47:00.437434  4610 caffe.cpp:169] Batch 1, accuracy = 0.99
I1203 18:47:00.437471  4610 caffe.cpp:169] Batch 1, loss = 0.0117609
I1203 18:47:00.439000  4610 caffe.cpp:169] Batch 2, accuracy = 1
I1203 18:47:00.439020  4610 caffe.cpp:169] Batch 2, loss = 0.00555977
I1203 18:47:00.440551  4610 caffe.cpp:169] Batch 3, accuracy = 0.99
I1203 18:47:00.440575  4610 caffe.cpp:169] Batch 3, loss = 0.0412139
I1203 18:47:00.442105  4610 caffe.cpp:169] Batch 4, accuracy = 0.99
I1203 18:47:00.442126  4610 caffe.cpp:169] Batch 4, loss = 0.0579313
I1203 18:47:00.443619  4610 caffe.cpp:169] Batch 5, accuracy = 0.99
I1203 18:47:00.443639  4610 caffe.cpp:169] Batch 5, loss = 0.0479742
I1203 18:47:00.445159  4610 caffe.cpp:169] Batch 6, accuracy = 0.98
I1203 18:47:00.445179  4610 caffe.cpp:169] Batch 6, loss = 0.0570176
I1203 18:47:00.446712  4610 caffe.cpp:169] Batch 7, accuracy = 0.99
I1203 18:47:00.446732  4610 caffe.cpp:169] Batch 7, loss = 0.0272363
I1203 18:47:00.448249  4610 caffe.cpp:169] Batch 8, accuracy = 1
I1203 18:47:00.448269  4610 caffe.cpp:169] Batch 8, loss = 0.00680142
I1203 18:47:00.449801  4610 caffe.cpp:169] Batch 9, accuracy = 0.98
I1203 18:47:00.449821  4610 caffe.cpp:169] Batch 9, loss = 0.0288398
I1203 18:47:00.451352  4610 caffe.cpp:169] Batch 10, accuracy = 0.98
I1203 18:47:00.451372  4610 caffe.cpp:169] Batch 10, loss = 0.0603264
I1203 18:47:00.452883  4610 caffe.cpp:169] Batch 11, accuracy = 0.98
I1203 18:47:00.452903  4610 caffe.cpp:169] Batch 11, loss = 0.0524943
I1203 18:47:00.454407  4610 caffe.cpp:169] Batch 12, accuracy = 0.95
I1203 18:47:00.454427  4610 caffe.cpp:169] Batch 12, loss = 0.106648
I1203 18:47:00.455955  4610 caffe.cpp:169] Batch 13, accuracy = 0.98
I1203 18:47:00.455976  4610 caffe.cpp:169] Batch 13, loss = 0.0450225
I1203 18:47:00.457484  4610 caffe.cpp:169] Batch 14, accuracy = 1
I1203 18:47:00.457504  4610 caffe.cpp:169] Batch 14, loss = 0.00531614
I1203 18:47:00.459038  4610 caffe.cpp:169] Batch 15, accuracy = 0.98
I1203 18:47:00.459056  4610 caffe.cpp:169] Batch 15, loss = 0.065209
I1203 18:47:00.460577  4610 caffe.cpp:169] Batch 16, accuracy = 0.98
I1203 18:47:00.460597  4610 caffe.cpp:169] Batch 16, loss = 0.0520317
I1203 18:47:00.462123  4610 caffe.cpp:169] Batch 17, accuracy = 0.99
I1203 18:47:00.462143  4610 caffe.cpp:169] Batch 17, loss = 0.0328681
I1203 18:47:00.463656  4610 caffe.cpp:169] Batch 18, accuracy = 0.99
I1203 18:47:00.463676  4610 caffe.cpp:169] Batch 18, loss = 0.0175973
I1203 18:47:00.465188  4610 caffe.cpp:169] Batch 19, accuracy = 0.97
I1203 18:47:00.465208  4610 caffe.cpp:169] Batch 19, loss = 0.0576884
I1203 18:47:00.466749  4610 caffe.cpp:169] Batch 20, accuracy = 0.97
I1203 18:47:00.466769  4610 caffe.cpp:169] Batch 20, loss = 0.0850501
I1203 18:47:00.468278  4610 caffe.cpp:169] Batch 21, accuracy = 0.98
I1203 18:47:00.468298  4610 caffe.cpp:169] Batch 21, loss = 0.0676049
I1203 18:47:00.469805  4610 caffe.cpp:169] Batch 22, accuracy = 0.99
I1203 18:47:00.469825  4610 caffe.cpp:169] Batch 22, loss = 0.0448538
I1203 18:47:00.471328  4610 caffe.cpp:169] Batch 23, accuracy = 0.97
I1203 18:47:00.471349  4610 caffe.cpp:169] Batch 23, loss = 0.0333992
I1203 18:47:00.487124  4610 caffe.cpp:169] Batch 24, accuracy = 1
I1203 18:47:00.487180  4610 caffe.cpp:169] Batch 24, loss = 0.0281527
I1203 18:47:00.489002  4610 caffe.cpp:169] Batch 25, accuracy = 0.99
I1203 18:47:00.489048  4610 caffe.cpp:169] Batch 25, loss = 0.0545881
I1203 18:47:00.490890  4610 caffe.cpp:169] Batch 26, accuracy = 0.98
I1203 18:47:00.490932  4610 caffe.cpp:169] Batch 26, loss = 0.115576
I1203 18:47:00.492620  4610 caffe.cpp:169] Batch 27, accuracy = 1
I1203 18:47:00.492640  4610 caffe.cpp:169] Batch 27, loss = 0.0149555
I1203 18:47:00.494161  4610 caffe.cpp:169] Batch 28, accuracy = 0.98
I1203 18:47:00.494181  4610 caffe.cpp:169] Batch 28, loss = 0.0398991
I1203 18:47:00.495693  4610 caffe.cpp:169] Batch 29, accuracy = 0.96
I1203 18:47:00.495713  4610 caffe.cpp:169] Batch 29, loss = 0.115862
I1203 18:47:00.497226  4610 caffe.cpp:169] Batch 30, accuracy = 1
I1203 18:47:00.497246  4610 caffe.cpp:169] Batch 30, loss = 0.0116793
I1203 18:47:00.498785  4610 caffe.cpp:169] Batch 31, accuracy = 1
I1203 18:47:00.498817  4610 caffe.cpp:169] Batch 31, loss = 0.00451814
I1203 18:47:00.500329  4610 caffe.cpp:169] Batch 32, accuracy = 0.98
I1203 18:47:00.500349  4610 caffe.cpp:169] Batch 32, loss = 0.0244668
I1203 18:47:00.501878  4610 caffe.cpp:169] Batch 33, accuracy = 1
I1203 18:47:00.501899  4610 caffe.cpp:169] Batch 33, loss = 0.00285445
I1203 18:47:00.503411  4610 caffe.cpp:169] Batch 34, accuracy = 0.98
I1203 18:47:00.503429  4610 caffe.cpp:169] Batch 34, loss = 0.0566256
I1203 18:47:00.504940  4610 caffe.cpp:169] Batch 35, accuracy = 0.95
I1203 18:47:00.504961  4610 caffe.cpp:169] Batch 35, loss = 0.154924
I1203 18:47:00.506500  4610 caffe.cpp:169] Batch 36, accuracy = 1
I1203 18:47:00.506520  4610 caffe.cpp:169] Batch 36, loss = 0.00451233
I1203 18:47:00.508111  4610 caffe.cpp:169] Batch 37, accuracy = 0.97
I1203 18:47:00.508131  4610 caffe.cpp:169] Batch 37, loss = 0.0572309
I1203 18:47:00.509635  4610 caffe.cpp:169] Batch 38, accuracy = 0.99
I1203 18:47:00.509655  4610 caffe.cpp:169] Batch 38, loss = 0.0192229
I1203 18:47:00.511181  4610 caffe.cpp:169] Batch 39, accuracy = 0.99
I1203 18:47:00.511200  4610 caffe.cpp:169] Batch 39, loss = 0.029272
I1203 18:47:00.512725  4610 caffe.cpp:169] Batch 40, accuracy = 0.99
I1203 18:47:00.512745  4610 caffe.cpp:169] Batch 40, loss = 0.0258552
I1203 18:47:00.514317  4610 caffe.cpp:169] Batch 41, accuracy = 0.99
I1203 18:47:00.514338  4610 caffe.cpp:169] Batch 41, loss = 0.0752082
I1203 18:47:00.515854  4610 caffe.cpp:169] Batch 42, accuracy = 1
I1203 18:47:00.515873  4610 caffe.cpp:169] Batch 42, loss = 0.0283319
I1203 18:47:00.517379  4610 caffe.cpp:169] Batch 43, accuracy = 0.99
I1203 18:47:00.517398  4610 caffe.cpp:169] Batch 43, loss = 0.0112394
I1203 18:47:00.518925  4610 caffe.cpp:169] Batch 44, accuracy = 0.98
I1203 18:47:00.518946  4610 caffe.cpp:169] Batch 44, loss = 0.0413653
I1203 18:47:00.520457  4610 caffe.cpp:169] Batch 45, accuracy = 0.98
I1203 18:47:00.520478  4610 caffe.cpp:169] Batch 45, loss = 0.0501227
I1203 18:47:00.521989  4610 caffe.cpp:169] Batch 46, accuracy = 1
I1203 18:47:00.522009  4610 caffe.cpp:169] Batch 46, loss = 0.0114459
I1203 18:47:00.523540  4610 caffe.cpp:169] Batch 47, accuracy = 1
I1203 18:47:00.523561  4610 caffe.cpp:169] Batch 47, loss = 0.0163504
I1203 18:47:00.525075  4610 caffe.cpp:169] Batch 48, accuracy = 0.97
I1203 18:47:00.525095  4610 caffe.cpp:169] Batch 48, loss = 0.0450363
I1203 18:47:00.526633  4610 caffe.cpp:169] Batch 49, accuracy = 1
I1203 18:47:00.526651  4610 caffe.cpp:169] Batch 49, loss = 0.0046898
I1203 18:47:00.526662  4610 caffe.cpp:174] Loss: 0.041468
I1203 18:47:00.526674  4610 caffe.cpp:186] accuracy = 0.9856
I1203 18:47:00.526687  4610 caffe.cpp:186] loss = 0.041468 (* 1 = 0.041468 loss)


你可能感兴趣的:(神经网络,cnn,深度学习,caffe)