【深度学习】学习笔记：Caffe上配置和运行Cifar10的示例

最新推荐文章于 2022-06-06 18:46:37 发布
最新推荐文章于 2022-06-06 18:46:37 发布
阅读量472 收藏
点赞数

CIFAR-10数据集含有6万个32*32的彩色图像，共分为10种类型，由 Alex Krizhevsky, Vinod Nair和 Geoffrey Hinton收集而来。包含50000张训练图片，10000张测试图片

http://www.cs.toronto.edu/~kriz/cifar.html

数据集的数据存在一个10000*3072 的 numpy数组中，单位是uint8s，3072是存储了一个32*32的彩色图像。（3072=1024*3=32*32*3）。前1024位是r值，中间1024是g值，后面1024是b值(RGB)。



首先准备数据集：


# /home/nvidia/caffe/data/cifar10

cd $CAFFE_ROOT/data/cifar10

./get_cifar10.sh



:~/caffe$ cd /home/nvidia/caffe/data/cifar10

nvidia@tegra-ubuntu:~/caffe$ ./get_cifar10.sh


[html]  view plain  copy
  1. [root@localhost cifar10]# ./get_cifar10.sh   
  2. Downloading...  
  3. --2015-03-09 18:38:23--  http://www.cs.toronto.edu/~kriz/cifar-10-binary.tar.gz  
  4. Resolving www.cs.toronto.edu... 128.100.3.30  
  5. Connecting to www.cs.toronto.edu|128.100.3.30|:80... connected.  
  6. HTTP request sent, awaiting response... 200 OK  
  7. Length: 170052171 (162M) [application/x-gzip]  
  8. Saving to: 鈥渃ifar-10-binary.tar.gz鈥  
  9.   
  10. 100%[=============================================================>] 170,052,171 1.66M/s   in 5m 22s    
[html]  view plain  copy
  1. Unzipping...  
  2. Done.  



cifar-10-binary.tar.gz   100%[++===============================>] 162.17M  16.8KB/s    in 3h 2m   
2018-01-15 19:59:39 (14.1 KB/s) - ‘cifar-10-binary.tar.gz’ saved [170052171/170052171]
Unzipping...
Done.


然后进入目录 

执行cd $CAFFE_ROOT/examples/cifar10

./create_cifar10.sh


nvidia@tegra-ubuntu:~/caffe$ ./examples/cifar10/create_cifar10.shcreate_cifar10.s

源代码如下，我们可以看到，所做的工作就是将图片库转成leveldb格式，并计算均值二进制文件

[html]  view plain  copy
  1. #!/usr/bin/env sh  
  2. # This script converts the cifar data into leveldb format.  
  3.   
  4. EXAMPLE=examples/cifar10  
  5. DATA=data/cifar10  
  6.   
  7. echo "Creating leveldb..."  
  8.   
  9. rm -rf $EXAMPLE/cifar10_train_leveldb $EXAMPLE/cifar10_test_leveldb  
  10.   
  11. ./build/examples/cifar10/convert_cifar_data.bin $DATA $EXAMPLE  
  12.   
  13. echo "Computing image mean..."  
  14.   
  15. ./build/tools/compute_image_mean $EXAMPLE/cifar10_train_leveldb \  
  16.   $EXAMPLE/mean.binaryproto leveldb  
  17.   
  18. echo "Done."  

#!/usr/bin/env sh
# This script converts the cifar data into leveldb format.
set -e

EXAMPLE=examples/cifar10
DATA=data/cifar10
DBTYPE=lmdb

echo "Creating $DBTYPE..."

rm -rf $EXAMPLE/cifar10_train_$DBTYPE $EXAMPLE/cifar10_test_$DBTYPE

./build/examples/cifar10/convert_cifar_data.bin $DATA $EXAMPLE $DBTYPE

echo "Computing image mean..."

./build/tools/compute_image_mean -backend=$DBTYPE \
  $EXAMPLE/cifar10_train_$DBTYPE $EXAMPLE/mean.binaryproto

echo "Done."




运行之后，将会在 examples 中出现数据库文件 ./cifar10-leveldb 和数据库图像均值二进制文件 ./mean.binaryproto
[html]  view plain  copy
  1. [root@localhost caffe]# ./examples/cifar10/create_cifar10.sh  
  2. Creating leveldb...  
  3. Computing image mean...  
  4. E0309 18:50:12.026103 18241 compute_image_mean.cpp:114] Processed 10000 files.  
  5. E0309 18:50:12.094758 18241 compute_image_mean.cpp:114] Processed 20000 files.  
  6. E0309 18:50:12.163048 18241 compute_image_mean.cpp:114] Processed 30000 files.  
  7. E0309 18:50:12.231233 18241 compute_image_mean.cpp:114] Processed 40000 files.  
  8. E0309 18:50:12.299324 18241 compute_image_mean.cpp:114] Processed 50000 files.  
  9. Done.  


模型描述在examples/cifar10/cifar10_quick_solver.prototxt，和examples/cifar10/cifar10_quick_train_test.prototxt 中。



[html]  view plain  copy
  1. # reduce the learning rate after 8 epochs (4000 iters) by a factor of 10  
  2.   
  3. # The train/test net protocol buffer definition  
  4. net: "examples/cifar10/cifar10_quick_train_test.prototxt"  
  5. # test_iter specifies how many forward passes the test should carry out.  
  6. # In the case of MNIST, we have test batch size 100 and 100 test iterations,  
  7. # covering the full 10,000 testing images.  
  8. test_iter: 100  
  9. # Carry out testing every 500 training iterations.  
  10. test_interval: 500  
  11. # The base learning rate, momentum and the weight decay of the network.  
  12. base_lr: 0.001  
  13. momentum: 0.9  
  14. weight_decay: 0.004  
  15. # The learning rate policy  
  16. lr_policy: "fixed"  
  17. # Display every 100 iterations  
  18. display: 100  
  19. # The maximum number of iterations  
  20. max_iter: 4000  
  21. # snapshot intermediate results  
  22. snapshot: 4000  
  23. snapshot_prefix: "examples/cifar10/cifar10_quick"  
  24. # solver mode: CPU or GPU  
  25. solver_mode: CPU  
  26. ~  


[html]  view plain  copy
  1. name: "CIFAR10_quick"  
  2. layers {  
  3.   name: "cifar"  
  4.   type: DATA  
  5.   top: "data"  
  6.   top: "label"  
  7.   data_param {  
  8.     source: "examples/cifar10/cifar10_train_leveldb"  
  9.     batch_size: 100  
  10.   }  
  11.   transform_param {  
  12.     mean_file: "examples/cifar10/mean.binaryproto"  
  13.   }  
  14.   include: { phase: TRAIN }  
  15. }  
  16. layers {  
  17.   name: "cifar"  
  18.   type: DATA  
  19.   top: "data"  
  20.   top: "label"  
  21.   data_param {  
  22.     source: "examples/cifar10/cifar10_test_leveldb"  
  23.     batch_size: 100  
  24.   }  
  25.   transform_param {  
  26.     mean_file: "examples/cifar10/mean.binaryproto"  
  27.   }  
  28.   include: { phase: TEST }  
  29. }  
  30. layers {  
  31.   name: "conv1"  
  32.   type: CONVOLUTION  
  33.   bottom: "data"  
  34.   top: "conv1"  
  35.   blobs_lr: 1  
  36.   blobs_lr: 2  
  37.   convolution_param {  
  38.     num_output: 32  
  39.     pad: 2  
  40.     kernel_size: 5  
  41.     stride: 1  
  42.     weight_filler {  
  43.       type: "gaussian"  
  44.       std: 0.0001  
  45.     }  
  46.     bias_filler {  
  47.       type: "constant"  
  48.     }  
  49.   }  
  50. }  
  51. layers {  
  52.   name: "pool1"  
  53.   type: POOLING  
  54.   bottom: "conv1"  
  55.   top: "pool1"  
  56.   pooling_param {  
  57.     pool: MAX  
  58.     kernel_size: 3  
  59.     stride: 2  
  60.   }  
  61. }  
  62. layers {  
  63.   name: "relu1"  
  64.   type: RELU  
  65.   bottom: "pool1"  
  66.   top: "pool1"  
  67. }  
  68. layers {  
  69.   name: "conv2"  
  70.   type: CONVOLUTION  
  71.   bottom: "pool1"  
  72.   top: "conv2"  
  73.   blobs_lr: 1  
  74.   blobs_lr: 2  
  75.   convolution_param {  
  76.     num_output: 32  
  77.     pad: 2  
  78.     kernel_size: 5  
  79.     stride: 1  
  80.     weight_filler {  
  81.       type: "gaussian"  
  82.       std: 0.01  
  83.     }  
  84.     bias_filler {  
  85.       type: "constant"  
  86.     }  
  87.   }  
  88. }  
  89. layers {  
  90.   name: "relu2"  
  91.   type: RELU  
  92.   bottom: "conv2"  
  93.   top: "conv2"  
  94. }  
  95. layers {  
  96.   name: "pool2"  
  97.   type: POOLING  
  98.   bottom: "conv2"  
  99.   top: "pool2"  
  100.   pooling_param {  
  101.     pool: AVE  
  102.     kernel_size: 3  
  103.     stride: 2  
  104.   }  
  105. }  
  106. layers {  
  107.   name: "conv3"  
  108.   type: CONVOLUTION  
  109.   bottom: "pool2"  
  110.   top: "conv3"  
  111.   blobs_lr: 1  
  112.   blobs_lr: 2  
  113.   convolution_param {  
  114.     num_output: 64  
  115.     pad: 2  
  116.     kernel_size: 5  
  117.     stride: 1  
  118.     weight_filler {  
  119.       type: "gaussian"  
  120.       std: 0.01  
  121.     }  
  122.     bias_filler {  
  123.       type: "constant"  
  124.     }  
  125.   }  
  126. }  
  127. layers {  
  128.   name: "relu3"  
  129.   type: RELU  
  130.   bottom: "conv3"  
  131.   top: "conv3"  
  132. }  
  133. layers {  
  134.   name: "pool3"  
  135.   type: POOLING  
  136.   bottom: "conv3"  
  137.   top: "pool3"  
  138.   pooling_param {  
  139.     pool: AVE  
  140.     kernel_size: 3  
  141.     stride: 2  
  142.   }  
  143. }  
  144. layers {  
  145.   name: "ip1"  
  146.   type: INNER_PRODUCT  
  147.   bottom: "pool3"  
  148.   top: "ip1"  
  149.   blobs_lr: 1  
  150.   blobs_lr: 2  
  151.   inner_product_param {  
  152.     num_output: 64  
  153.     weight_filler {  
  154.       type: "gaussian"  
  155.       std: 0.1  
  156.     }  
  157.     bias_filler {  
  158.       type: "constant"  
  159.     }  
  160.   }  
  161. }  
  162. layers {  
  163.   name: "ip2"  
  164.   type: INNER_PRODUCT  
  165.   bottom: "ip1"  
  166.   top: "ip2"  
  167.   blobs_lr: 1  
  168.   blobs_lr: 2  
  169.   inner_product_param {  
  170.     num_output: 10  
  171.     weight_filler {  
  172.       type: "gaussian"  
  173.       std: 0.1  
  174.     }  
  175.     bias_filler {  
  176.       type: "constant"  
  177.     }  
  178.   }  
  179. }  
  180. layers {  
  181.   name: "accuracy"  
  182.   type: ACCURACY  
  183.   bottom: "ip2"  
  184.   bottom: "label"  
  185.   top: "accuracy"  
  186.   include: { phase: TEST }  
  187. }  
  188. layers {  
  189.   name: "loss"  
  190.   type: SOFTMAX_LOSS  
  191.   bottom: "ip2"  
  192.   bottom: "label"  
  193.   top: "loss"  
  194. }  




模型训练是 执行 train_quick.sh ，内容如下

[html]  view plain  copy
  1. [root@localhost cifar10]# vi train_quick.sh   
  2.   
  3.   
  4. #!/usr/bin/env sh  
  5.   
  6.   
  7. TOOLS=./build/tools  
  8.   
  9.   
  10. $TOOLS/caffe train \  
  11.   --solver=examples/cifar10/cifar10_quick_solver.prototxt  
  12.   
  13.   
  14. # reduce learning rate by factor of 10 after 8 epochs  
  15. $TOOLS/caffe train \  
  16.   --solver=examples/cifar10/cifar10_quick_solver_lr1.prototxt \  
  17.   --snapshot=examples/cifar10/cifar10_quick_iter_4000.solverstate  
  18. ~  
执行结果如下：

  1.       
  2. cifar10_quick_iter_5000.solverstate  
  3. I0317 22:12:25.592813 2008298256 solver.cpp:81] Optimization Done.  
  4. Our model achieved ~75% test accuracy. The model parameters are stored in binary protobuf format in  
  5.   
  6. cifar10_quick_iter_5000  


nvidia@tegra-ubuntu:~/caffe/examples/cifar10$ train_quick.sh
bash: train_quick.sh: command not found
nvidia@tegra-ubuntu:~/caffe/examples/cifar10$ ./train_quick.sh
./train_quick.sh: 6: ./train_quick.sh: ./build/tools/caffe: not found
nvidia@tegra-ubuntu:~/caffe/examples/cifar10$ 

caffe训练cifar10遇到./build/tools/caffe: not found 错误解决方法

cifar10训练步骤如下：

（1）打开终端，应用cd切换路径，如 cd ~/caffe/data/cifar10 ，

（2）继续执行命令   ./get_cifar10.sh，

（3）成功下载数据集之后，执行ls即可见所下载的数据文件，

（4）再次将路径切换到cd ~/caffe/examples/cifar10

（5）继续执行命令 ./create_cifar10.sh 

此时系统报错./build/examples/cifar10/convert_cifar_data.bin: not found ，这是因为当前目录在/caffe/examples/cifar10 ，而执行./create_cifar10.sh ，需要在caffe目录下，因此我们需要切换到caffe目录下，然后执行./examples/cifar10/create_cifar10.sh 

(6) 同理，在caffe目录下执行./examples/cifar10/train_quick.sh ,此时就不会报错找不到 ./build/tools/caffe 了。


执行结果如下：

nvidia@tegra-ubuntu:~/caffe$ ./examples/cifar10/train_quick.sh
I0116 13:04:53.104024 25452 caffe.cpp:218] Using GPUs 0
I0116 13:04:53.111429 25452 caffe.cpp:223] GPU 0: NVIDIA Tegra X2
I0116 13:04:53.700456 25452 solver.cpp:44] Initializing solver from parameters: 
test_iter: 100
test_interval: 500
base_lr: 0.001
display: 100
max_iter: 4000
lr_policy: "fixed"
momentum: 0.9
weight_decay: 0.004
snapshot: 4000
snapshot_prefix: "examples/cifar10/cifar10_quick"
solver_mode: GPU
device_id: 0
net: "examples/cifar10/cifar10_quick_train_test.prototxt"
train_state {
  level: 0
  stage: ""
}
I0116 13:04:53.700881 25452 solver.cpp:87] Creating training net from net file: examples/cifar10/cifar10_quick_train_test.prototxt
I0116 13:04:53.701413 25452 net.cpp:294] The NetState phase (0) differed from the phase (1) specified by a rule in layer cifar
I0116 13:04:53.701463 25452 net.cpp:294] The NetState phase (0) differed from the phase (1) specified by a rule in layer accuracy
I0116 13:04:53.701498 25452 net.cpp:51] Initializing net from parameters: 
name: "CIFAR10_quick"
state {
  phase: TRAIN
  level: 0
  stage: ""
}
layer {
  name: "cifar"
  type: "Data"
  top: "data"
  top: "label"
  include {
    phase: TRAIN
  }
  transform_param {
    mean_file: "examples/cifar10/mean.binaryproto"
  }
  data_param {
    source: "examples/cifar10/cifar10_train_lmdb"
    batch_size: 100
    backend: LMDB
  }
}
layer {
  name: "conv1"
  type: "Convolution"
  bottom: "data"
  top: "conv1"
  param {
    lr_mult: 1
  }
  param {
    lr_mult: 2
  }
  convolution_param {
    num_output: 32
    pad: 2
    kernel_size: 5
    stride: 1
    weight_filler {
      type: "gaussian"
      std: 0.0001
    }
    bias_filler {
      type: "constant"
    }
  }
}
layer {
  name: "pool1"
  type: "Pooling"
  bottom: "conv1"
  top: "pool1"
  pooling_param {
    pool: MAX
    kernel_size: 3
    stride: 2
  }
}
layer {
  name: "relu1"
  type: "ReLU"
  bottom: "pool1"
  top: "pool1"
}
layer {
  name: "conv2"
  type: "Convolution"
  bottom: "pool1"
  top: "conv2"
  param {
    lr_mult: 1
  }
  param {
    lr_mult: 2
  }
  convolution_param {
    num_output: 32
    pad: 2
    kernel_size: 5
    stride: 1
    weight_filler {
      type: "gaussian"
      std: 0.01
    }
    bias_filler {
      type: "constant"
    }
  }
}
layer {
  name: "relu2"
  type: "ReLU"
  bottom: "conv2"
  top: "conv2"
}
layer {
  name: "pool2"
  type: "Pooling"
  bottom: "conv2"
  top: "pool2"
  pooling_param {
    pool: AVE
    kernel_size: 3
    stride: 2
  }
}
layer {
  name: "conv3"
  type: "Convolution"
  bottom: "pool2"
  top: "conv3"
  param {
    lr_mult: 1
  }
  param {
    lr_mult: 2
  }
  convolution_param {
    num_output: 64
    pad: 2
    kernel_size: 5
    stride: 1
    weight_filler {
      type: "gaussian"
      std: 0.01
    }
    bias_filler {
      type: "constant"
    }
  }
}
layer {
  name: "relu3"
  type: "ReLU"
  bottom: "conv3"
  top: "conv3"
}
layer {
  name: "pool3"
  type: "Pooling"
  bottom: "conv3"
  top: "pool3"
  pooling_param {
    pool: AVE
    kernel_size: 3
    stride: 2
  }
}
layer {
  name: "ip1"
  type: "InnerProduct"
  bottom: "pool3"
  top: "ip1"
  param {
    lr_mult: 1
  }
  param {
    lr_mult: 2
  }
  inner_product_param {
    num_output: 64
    weight_filler {
      type: "gaussian"
      std: 0.1
    }
    bias_filler {
      type: "constant"
    }
  }
}
layer {
  name: "ip2"
  type: "InnerProduct"
  bottom: "ip1"
  top: "ip2"
  param {
    lr_mult: 1
  }
  param {
    lr_mult: 2
  }
  inner_product_param {
    num_output: 10
    weight_filler {
      type: "gaussian"
      std: 0.1
    }
    bias_filler {
      type: "constant"
    }
  }
}
layer {
  name: "loss"
  type: "SoftmaxWithLoss"
  bottom: "ip2"
  bottom: "label"
  top: "loss"
}
I0116 13:04:53.702026 25452 layer_factory.hpp:77] Creating layer cifar
I0116 13:04:53.702291 25452 db_lmdb.cpp:35] Opened lmdb examples/cifar10/cifar10_train_lmdb
I0116 13:04:53.702373 25452 net.cpp:84] Creating Layer cifar
I0116 13:04:53.702400 25452 net.cpp:380] cifar -> data
I0116 13:04:53.702452 25452 net.cpp:380] cifar -> label
I0116 13:04:53.702494 25452 data_transformer.cpp:25] Loading mean file from: examples/cifar10/mean.binaryproto
I0116 13:04:53.703125 25452 data_layer.cpp:45] output data size: 100,3,32,32
I0116 13:04:53.712062 25452 net.cpp:122] Setting up cifar
I0116 13:04:53.712122 25452 net.cpp:129] Top shape: 100 3 32 32 (307200)
I0116 13:04:53.712155 25452 net.cpp:129] Top shape: 100 (100)
I0116 13:04:53.712179 25452 net.cpp:137] Memory required for data: 1229200
I0116 13:04:53.712219 25452 layer_factory.hpp:77] Creating layer conv1
I0116 13:04:53.712278 25452 net.cpp:84] Creating Layer conv1
I0116 13:04:53.712309 25452 net.cpp:406] conv1 <- data
I0116 13:04:53.712357 25452 net.cpp:380] conv1 -> conv1
I0116 13:04:55.004492 25452 net.cpp:122] Setting up conv1
I0116 13:04:55.004555 25452 net.cpp:129] Top shape: 100 32 32 32 (3276800)
I0116 13:04:55.004581 25452 net.cpp:137] Memory required for data: 14336400
I0116 13:04:55.004644 25452 layer_factory.hpp:77] Creating layer pool1
I0116 13:04:55.004680 25452 net.cpp:84] Creating Layer pool1
I0116 13:04:55.004698 25452 net.cpp:406] pool1 <- conv1
I0116 13:04:55.004719 25452 net.cpp:380] pool1 -> pool1
I0116 13:04:55.004884 25452 net.cpp:122] Setting up pool1
I0116 13:04:55.004904 25452 net.cpp:129] Top shape: 100 32 16 16 (819200)
I0116 13:04:55.004921 25452 net.cpp:137] Memory required for data: 17613200
I0116 13:04:55.004936 25452 layer_factory.hpp:77] Creating layer relu1
I0116 13:04:55.004956 25452 net.cpp:84] Creating Layer relu1
I0116 13:04:55.004969 25452 net.cpp:406] relu1 <- pool1
I0116 13:04:55.004987 25452 net.cpp:367] relu1 -> pool1 (in-place)
I0116 13:04:55.006839 25452 net.cpp:122] Setting up relu1
I0116 13:04:55.006880 25452 net.cpp:129] Top shape: 100 32 16 16 (819200)
I0116 13:04:55.006901 25452 net.cpp:137] Memory required for data: 20890000
I0116 13:04:55.006917 25452 layer_factory.hpp:77] Creating layer conv2
I0116 13:04:55.006959 25452 net.cpp:84] Creating Layer conv2
I0116 13:04:55.006978 25452 net.cpp:406] conv2 <- pool1
I0116 13:04:55.006999 25452 net.cpp:380] conv2 -> conv2
I0116 13:04:55.016883 25452 net.cpp:122] Setting up conv2
I0116 13:04:55.016943 25452 net.cpp:129] Top shape: 100 32 16 16 (819200)
I0116 13:04:55.016966 25452 net.cpp:137] Memory required for data: 24166800
I0116 13:04:55.017004 25452 layer_factory.hpp:77] Creating layer relu2
I0116 13:04:55.017032 25452 net.cpp:84] Creating Layer relu2
I0116 13:04:55.017048 25452 net.cpp:406] relu2 <- conv2
I0116 13:04:55.017068 25452 net.cpp:367] relu2 -> conv2 (in-place)
I0116 13:04:55.019407 25452 net.cpp:122] Setting up relu2
I0116 13:04:55.019454 25452 net.cpp:129] Top shape: 100 32 16 16 (819200)
I0116 13:04:55.019479 25452 net.cpp:137] Memory required for data: 27443600
I0116 13:04:55.019498 25452 layer_factory.hpp:77] Creating layer pool2
I0116 13:04:55.019531 25452 net.cpp:84] Creating Layer pool2
I0116 13:04:55.019546 25452 net.cpp:406] pool2 <- conv2
I0116 13:04:55.019567 25452 net.cpp:380] pool2 -> pool2
I0116 13:04:55.021615 25452 net.cpp:122] Setting up pool2
I0116 13:04:55.021663 25452 net.cpp:129] Top shape: 100 32 8 8 (204800)
I0116 13:04:55.021689 25452 net.cpp:137] Memory required for data: 28262800
I0116 13:04:55.021705 25452 layer_factory.hpp:77] Creating layer conv3
I0116 13:04:55.021750 25452 net.cpp:84] Creating Layer conv3
I0116 13:04:55.021775 25452 net.cpp:406] conv3 <- pool2
I0116 13:04:55.021822 25452 net.cpp:380] conv3 -> conv3
I0116 13:04:55.030450 25452 net.cpp:122] Setting up conv3
I0116 13:04:55.030504 25452 net.cpp:129] Top shape: 100 64 8 8 (409600)
I0116 13:04:55.030531 25452 net.cpp:137] Memory required for data: 29901200
I0116 13:04:55.030571 25452 layer_factory.hpp:77] Creating layer relu3
I0116 13:04:55.030604 25452 net.cpp:84] Creating Layer relu3
I0116 13:04:55.030622 25452 net.cpp:406] relu3 <- conv3
I0116 13:04:55.030643 25452 net.cpp:367] relu3 -> conv3 (in-place)
I0116 13:04:55.033023 25452 net.cpp:122] Setting up relu3
I0116 13:04:55.033071 25452 net.cpp:129] Top shape: 100 64 8 8 (409600)
I0116 13:04:55.033094 25452 net.cpp:137] Memory required for data: 31539600
I0116 13:04:55.033110 25452 layer_factory.hpp:77] Creating layer pool3
I0116 13:04:55.033138 25452 net.cpp:84] Creating Layer pool3
I0116 13:04:55.033205 25452 net.cpp:406] pool3 <- conv3
I0116 13:04:55.033238 25452 net.cpp:380] pool3 -> pool3
I0116 13:04:55.035491 25452 net.cpp:122] Setting up pool3
I0116 13:04:55.035544 25452 net.cpp:129] Top shape: 100 64 4 4 (102400)
I0116 13:04:55.035567 25452 net.cpp:137] Memory required for data: 31949200
I0116 13:04:55.035585 25452 layer_factory.hpp:77] Creating layer ip1
I0116 13:04:55.035616 25452 net.cpp:84] Creating Layer ip1
I0116 13:04:55.035632 25452 net.cpp:406] ip1 <- pool3
I0116 13:04:55.035656 25452 net.cpp:380] ip1 -> ip1
I0116 13:04:55.037981 25452 net.cpp:122] Setting up ip1
I0116 13:04:55.038019 25452 net.cpp:129] Top shape: 100 64 (6400)
I0116 13:04:55.038040 25452 net.cpp:137] Memory required for data: 31974800
I0116 13:04:55.038066 25452 layer_factory.hpp:77] Creating layer ip2
I0116 13:04:55.038094 25452 net.cpp:84] Creating Layer ip2
I0116 13:04:55.038110 25452 net.cpp:406] ip2 <- ip1
I0116 13:04:55.038130 25452 net.cpp:380] ip2 -> ip2
I0116 13:04:55.038504 25452 net.cpp:122] Setting up ip2
I0116 13:04:55.038527 25452 net.cpp:129] Top shape: 100 10 (1000)
I0116 13:04:55.038544 25452 net.cpp:137] Memory required for data: 31978800
I0116 13:04:55.038575 25452 layer_factory.hpp:77] Creating layer loss
I0116 13:04:55.038599 25452 net.cpp:84] Creating Layer loss
I0116 13:04:55.038612 25452 net.cpp:406] loss <- ip2
I0116 13:04:55.038627 25452 net.cpp:406] loss <- label
I0116 13:04:55.038650 25452 net.cpp:380] loss -> loss
I0116 13:04:55.038686 25452 layer_factory.hpp:77] Creating layer loss
I0116 13:04:55.041414 25452 net.cpp:122] Setting up loss
I0116 13:04:55.041461 25452 net.cpp:129] Top shape: (1)
I0116 13:04:55.041484 25452 net.cpp:132]     with loss weight 1
I0116 13:04:55.041535 25452 net.cpp:137] Memory required for data: 31978804
I0116 13:04:55.041555 25452 net.cpp:198] loss needs backward computation.
I0116 13:04:55.041586 25452 net.cpp:198] ip2 needs backward computation.
I0116 13:04:55.041601 25452 net.cpp:198] ip1 needs backward computation.
I0116 13:04:55.041615 25452 net.cpp:198] pool3 needs backward computation.
I0116 13:04:55.041630 25452 net.cpp:198] relu3 needs backward computation.
I0116 13:04:55.041642 25452 net.cpp:198] conv3 needs backward computation.
I0116 13:04:55.041657 25452 net.cpp:198] pool2 needs backward computation.
I0116 13:04:55.041672 25452 net.cpp:198] relu2 needs backward computation.
I0116 13:04:55.041685 25452 net.cpp:198] conv2 needs backward computation.
I0116 13:04:55.041699 25452 net.cpp:198] relu1 needs backward computation.
I0116 13:04:55.041712 25452 net.cpp:198] pool1 needs backward computation.
I0116 13:04:55.041730 25452 net.cpp:198] conv1 needs backward computation.
I0116 13:04:55.041744 25452 net.cpp:200] cifar does not need backward computation.
I0116 13:04:55.041756 25452 net.cpp:242] This network produces output loss
I0116 13:04:55.041788 25452 net.cpp:255] Network initialization done.
I0116 13:04:55.042315 25452 solver.cpp:172] Creating test net (#0) specified by net file: examples/cifar10/cifar10_quick_train_test.prototxt
I0116 13:04:55.042404 25452 net.cpp:294] The NetState phase (1) differed from the phase (0) specified by a rule in layer cifar
I0116 13:04:55.042443 25452 net.cpp:51] Initializing net from parameters: 
name: "CIFAR10_quick"
state {
  phase: TEST
}
layer {
  name: "cifar"
  type: "Data"
  top: "data"
  top: "label"
  include {
    phase: TEST
  }
  transform_param {
    mean_file: "examples/cifar10/mean.binaryproto"
  }
  data_param {
    source: "examples/cifar10/cifar10_test_lmdb"
    batch_size: 100
    backend: LMDB
  }
}
layer {
  name: "conv1"
  type: "Convolution"
  bottom: "data"
  top: "conv1"
  param {
    lr_mult: 1
  }
  param {
    lr_mult: 2
  }
  convolution_param {
    num_output: 32
    pad: 2
    kernel_size: 5
    stride: 1
    weight_filler {
      type: "gaussian"
      std: 0.0001
    }
    bias_filler {
      type: "constant"
    }
  }
}
layer {
  name: "pool1"
  type: "Pooling"
  bottom: "conv1"
  top: "pool1"
  pooling_param {
    pool: MAX
    kernel_size: 3
    stride: 2
  }
}
layer {
  name: "relu1"
  type: "ReLU"
  bottom: "pool1"
  top: "pool1"
}
layer {
  name: "conv2"
  type: "Convolution"
  bottom: "pool1"
  top: "conv2"
  param {
    lr_mult: 1
  }
  param {
    lr_mult: 2
  }
  convolution_param {
    num_output: 32
    pad: 2
    kernel_size: 5
    stride: 1
    weight_filler {
      type: "gaussian"
      std: 0.01
    }
    bias_filler {
      type: "constant"
    }
  }
}
layer {
  name: "relu2"
  type: "ReLU"
  bottom: "conv2"
  top: "conv2"
}
layer {
  name: "pool2"
  type: "Pooling"
  bottom: "conv2"
  top: "pool2"
  pooling_param {
    pool: AVE
    kernel_size: 3
    stride: 2
  }
}
layer {
  name: "conv3"
  type: "Convolution"
  bottom: "pool2"
  top: "conv3"
  param {
    lr_mult: 1
  }
  param {
    lr_mult: 2
  }
  convolution_param {
    num_output: 64
    pad: 2
    kernel_size: 5
    stride: 1
    weight_filler {
      type: "gaussian"
      std: 0.01
    }
    bias_filler {
      type: "constant"
    }
  }
}
layer {
  name: "relu3"
  type: "ReLU"
  bottom: "conv3"
  top: "conv3"
}
layer {
  name: "pool3"
  type: "Pooling"
  bottom: "conv3"
  top: "pool3"
  pooling_param {
    pool: AVE
    kernel_size: 3
    stride: 2
  }
}
layer {
  name: "ip1"
  type: "InnerProduct"
  bottom: "pool3"
  top: "ip1"
  param {
    lr_mult: 1
  }
  param {
    lr_mult: 2
  }
  inner_product_param {
    num_output: 64
    weight_filler {
      type: "gaussian"
      std: 0.1
    }
    bias_filler {
      type: "constant"
    }
  }
}
layer {
  name: "ip2"
  type: "InnerProduct"
  bottom: "ip1"
  top: "ip2"
  param {
    lr_mult: 1
  }
  param {
    lr_mult: 2
  }
  inner_product_param {
    num_output: 10
    weight_filler {
      type: "gaussian"
      std: 0.1
    }
    bias_filler {
      type: "constant"
    }
  }
}
layer {
  name: "accuracy"
  type: "Accuracy"
  bottom: "ip2"
  bottom: "label"
  top: "accuracy"
  include {
    phase: TEST
  }
}
layer {
  name: "loss"
  type: "SoftmaxWithLoss"
  bottom: "ip2"
  bottom: "label"
  top: "loss"
}
I0116 13:04:55.042997 25452 layer_factory.hpp:77] Creating layer cifar
I0116 13:04:55.043182 25452 db_lmdb.cpp:35] Opened lmdb examples/cifar10/cifar10_test_lmdb
I0116 13:04:55.043238 25452 net.cpp:84] Creating Layer cifar
I0116 13:04:55.043267 25452 net.cpp:380] cifar -> data
I0116 13:04:55.043295 25452 net.cpp:380] cifar -> label
I0116 13:04:55.043323 25452 data_transformer.cpp:25] Loading mean file from: examples/cifar10/mean.binaryproto
I0116 13:04:55.043987 25452 data_layer.cpp:45] output data size: 100,3,32,32
I0116 13:04:55.070179 25452 net.cpp:122] Setting up cifar
I0116 13:04:55.070256 25452 net.cpp:129] Top shape: 100 3 32 32 (307200)
I0116 13:04:55.070305 25452 net.cpp:129] Top shape: 100 (100)
I0116 13:04:55.070324 25452 net.cpp:137] Memory required for data: 1229200
I0116 13:04:55.070346 25452 layer_factory.hpp:77] Creating layer label_cifar_1_split
I0116 13:04:55.070382 25452 net.cpp:84] Creating Layer label_cifar_1_split
I0116 13:04:55.070402 25452 net.cpp:406] label_cifar_1_split <- label
I0116 13:04:55.070427 25452 net.cpp:380] label_cifar_1_split -> label_cifar_1_split_0
I0116 13:04:55.070463 25452 net.cpp:380] label_cifar_1_split -> label_cifar_1_split_1
I0116 13:04:55.070688 25452 net.cpp:122] Setting up label_cifar_1_split
I0116 13:04:55.070716 25452 net.cpp:129] Top shape: 100 (100)
I0116 13:04:55.070736 25452 net.cpp:129] Top shape: 100 (100)
I0116 13:04:55.070750 25452 net.cpp:137] Memory required for data: 1230000
I0116 13:04:55.070762 25452 layer_factory.hpp:77] Creating layer conv1
I0116 13:04:55.070798 25452 net.cpp:84] Creating Layer conv1
I0116 13:04:55.070816 25452 net.cpp:406] conv1 <- data
I0116 13:04:55.070838 25452 net.cpp:380] conv1 -> conv1
I0116 13:04:55.090713 25452 net.cpp:122] Setting up conv1
I0116 13:04:55.090768 25452 net.cpp:129] Top shape: 100 32 32 32 (3276800)
I0116 13:04:55.090797 25452 net.cpp:137] Memory required for data: 14337200
I0116 13:04:55.090842 25452 layer_factory.hpp:77] Creating layer pool1
I0116 13:04:55.090878 25452 net.cpp:84] Creating Layer pool1
I0116 13:04:55.090893 25452 net.cpp:406] pool1 <- conv1
I0116 13:04:55.090960 25452 net.cpp:380] pool1 -> pool1
I0116 13:04:55.091177 25452 net.cpp:122] Setting up pool1
I0116 13:04:55.091202 25452 net.cpp:129] Top shape: 100 32 16 16 (819200)
I0116 13:04:55.091223 25452 net.cpp:137] Memory required for data: 17614000
I0116 13:04:55.091238 25452 layer_factory.hpp:77] Creating layer relu1
I0116 13:04:55.091264 25452 net.cpp:84] Creating Layer relu1
I0116 13:04:55.091279 25452 net.cpp:406] relu1 <- pool1
I0116 13:04:55.091295 25452 net.cpp:367] relu1 -> pool1 (in-place)
I0116 13:04:55.093794 25452 net.cpp:122] Setting up relu1
I0116 13:04:55.093854 25452 net.cpp:129] Top shape: 100 32 16 16 (819200)
I0116 13:04:55.093879 25452 net.cpp:137] Memory required for data: 20890800
I0116 13:04:55.093895 25452 layer_factory.hpp:77] Creating layer conv2
I0116 13:04:55.093940 25452 net.cpp:84] Creating Layer conv2
I0116 13:04:55.093960 25452 net.cpp:406] conv2 <- pool1
I0116 13:04:55.093992 25452 net.cpp:380] conv2 -> conv2
I0116 13:04:55.106508 25452 net.cpp:122] Setting up conv2
I0116 13:04:55.106557 25452 net.cpp:129] Top shape: 100 32 16 16 (819200)
I0116 13:04:55.106582 25452 net.cpp:137] Memory required for data: 24167600
I0116 13:04:55.106617 25452 layer_factory.hpp:77] Creating layer relu2
I0116 13:04:55.106650 25452 net.cpp:84] Creating Layer relu2
I0116 13:04:55.106668 25452 net.cpp:406] relu2 <- conv2
I0116 13:04:55.106689 25452 net.cpp:367] relu2 -> conv2 (in-place)
I0116 13:04:55.109305 25452 net.cpp:122] Setting up relu2
I0116 13:04:55.109350 25452 net.cpp:129] Top shape: 100 32 16 16 (819200)
I0116 13:04:55.109374 25452 net.cpp:137] Memory required for data: 27444400
I0116 13:04:55.109390 25452 layer_factory.hpp:77] Creating layer pool2
I0116 13:04:55.109421 25452 net.cpp:84] Creating Layer pool2
I0116 13:04:55.109439 25452 net.cpp:406] pool2 <- conv2
I0116 13:04:55.109468 25452 net.cpp:380] pool2 -> pool2
I0116 13:04:55.112432 25452 net.cpp:122] Setting up pool2
I0116 13:04:55.112485 25452 net.cpp:129] Top shape: 100 32 8 8 (204800)
I0116 13:04:55.112509 25452 net.cpp:137] Memory required for data: 28263600
I0116 13:04:55.112531 25452 layer_factory.hpp:77] Creating layer conv3
I0116 13:04:55.112586 25452 net.cpp:84] Creating Layer conv3
I0116 13:04:55.112607 25452 net.cpp:406] conv3 <- pool2
I0116 13:04:55.112632 25452 net.cpp:380] conv3 -> conv3
I0116 13:04:55.126611 25452 net.cpp:122] Setting up conv3
I0116 13:04:55.126663 25452 net.cpp:129] Top shape: 100 64 8 8 (409600)
I0116 13:04:55.126687 25452 net.cpp:137] Memory required for data: 29902000
I0116 13:04:55.126722 25452 layer_factory.hpp:77] Creating layer relu3
I0116 13:04:55.126756 25452 net.cpp:84] Creating Layer relu3
I0116 13:04:55.126772 25452 net.cpp:406] relu3 <- conv3
I0116 13:04:55.126794 25452 net.cpp:367] relu3 -> conv3 (in-place)
I0116 13:04:55.129612 25452 net.cpp:122] Setting up relu3
I0116 13:04:55.129658 25452 net.cpp:129] Top shape: 100 64 8 8 (409600)
I0116 13:04:55.129683 25452 net.cpp:137] Memory required for data: 31540400
I0116 13:04:55.129700 25452 layer_factory.hpp:77] Creating layer pool3
I0116 13:04:55.129739 25452 net.cpp:84] Creating Layer pool3
I0116 13:04:55.129760 25452 net.cpp:406] pool3 <- conv3
I0116 13:04:55.129786 25452 net.cpp:380] pool3 -> pool3
I0116 13:04:55.132236 25452 net.cpp:122] Setting up pool3
I0116 13:04:55.132287 25452 net.cpp:129] Top shape: 100 64 4 4 (102400)
I0116 13:04:55.132308 25452 net.cpp:137] Memory required for data: 31950000
I0116 13:04:55.132323 25452 layer_factory.hpp:77] Creating layer ip1
I0116 13:04:55.132359 25452 net.cpp:84] Creating Layer ip1
I0116 13:04:55.132374 25452 net.cpp:406] ip1 <- pool3
I0116 13:04:55.132397 25452 net.cpp:380] ip1 -> ip1
I0116 13:04:55.135041 25452 net.cpp:122] Setting up ip1
I0116 13:04:55.135087 25452 net.cpp:129] Top shape: 100 64 (6400)
I0116 13:04:55.135108 25452 net.cpp:137] Memory required for data: 31975600
I0116 13:04:55.135134 25452 layer_factory.hpp:77] Creating layer ip2
I0116 13:04:55.135164 25452 net.cpp:84] Creating Layer ip2
I0116 13:04:55.135180 25452 net.cpp:406] ip2 <- ip1
I0116 13:04:55.135215 25452 net.cpp:380] ip2 -> ip2
I0116 13:04:55.135763 25452 net.cpp:122] Setting up ip2
I0116 13:04:55.135787 25452 net.cpp:129] Top shape: 100 10 (1000)
I0116 13:04:55.135807 25452 net.cpp:137] Memory required for data: 31979600
I0116 13:04:55.135836 25452 layer_factory.hpp:77] Creating layer ip2_ip2_0_split
I0116 13:04:55.135859 25452 net.cpp:84] Creating Layer ip2_ip2_0_split
I0116 13:04:55.135872 25452 net.cpp:406] ip2_ip2_0_split <- ip2
I0116 13:04:55.135890 25452 net.cpp:380] ip2_ip2_0_split -> ip2_ip2_0_split_0
I0116 13:04:55.135911 25452 net.cpp:380] ip2_ip2_0_split -> ip2_ip2_0_split_1
I0116 13:04:55.136051 25452 net.cpp:122] Setting up ip2_ip2_0_split
I0116 13:04:55.136068 25452 net.cpp:129] Top shape: 100 10 (1000)
I0116 13:04:55.136085 25452 net.cpp:129] Top shape: 100 10 (1000)
I0116 13:04:55.136098 25452 net.cpp:137] Memory required for data: 31987600
I0116 13:04:55.136111 25452 layer_factory.hpp:77] Creating layer accuracy
I0116 13:04:55.136135 25452 net.cpp:84] Creating Layer accuracy
I0116 13:04:55.136157 25452 net.cpp:406] accuracy <- ip2_ip2_0_split_0
I0116 13:04:55.136171 25452 net.cpp:406] accuracy <- label_cifar_1_split_0
I0116 13:04:55.136193 25452 net.cpp:380] accuracy -> accuracy
I0116 13:04:55.136224 25452 net.cpp:122] Setting up accuracy
I0116 13:04:55.136236 25452 net.cpp:129] Top shape: (1)
I0116 13:04:55.136251 25452 net.cpp:137] Memory required for data: 31987604
I0116 13:04:55.136265 25452 layer_factory.hpp:77] Creating layer loss
I0116 13:04:55.136284 25452 net.cpp:84] Creating Layer loss
I0116 13:04:55.136297 25452 net.cpp:406] loss <- ip2_ip2_0_split_1
I0116 13:04:55.136314 25452 net.cpp:406] loss <- label_cifar_1_split_1
I0116 13:04:55.136335 25452 net.cpp:380] loss -> loss
I0116 13:04:55.136365 25452 layer_factory.hpp:77] Creating layer loss
I0116 13:04:55.139508 25452 net.cpp:122] Setting up loss
I0116 13:04:55.139556 25452 net.cpp:129] Top shape: (1)
I0116 13:04:55.139580 25452 net.cpp:132]     with loss weight 1
I0116 13:04:55.139606 25452 net.cpp:137] Memory required for data: 31987608
I0116 13:04:55.139624 25452 net.cpp:198] loss needs backward computation.
I0116 13:04:55.139642 25452 net.cpp:200] accuracy does not need backward computation.
I0116 13:04:55.139658 25452 net.cpp:198] ip2_ip2_0_split needs backward computation.
I0116 13:04:55.139672 25452 net.cpp:198] ip2 needs backward computation.
I0116 13:04:55.139686 25452 net.cpp:198] ip1 needs backward computation.
I0116 13:04:55.139699 25452 net.cpp:198] pool3 needs backward computation.
I0116 13:04:55.139716 25452 net.cpp:198] relu3 needs backward computation.
I0116 13:04:55.139729 25452 net.cpp:198] conv3 needs backward computation.
I0116 13:04:55.139744 25452 net.cpp:198] pool2 needs backward computation.
I0116 13:04:55.139757 25452 net.cpp:198] relu2 needs backward computation.
I0116 13:04:55.139770 25452 net.cpp:198] conv2 needs backward computation.
I0116 13:04:55.139784 25452 net.cpp:198] relu1 needs backward computation.
I0116 13:04:55.139801 25452 net.cpp:198] pool1 needs backward computation.
I0116 13:04:55.139816 25452 net.cpp:198] conv1 needs backward computation.
I0116 13:04:55.139830 25452 net.cpp:200] label_cifar_1_split does not need backward computation.
I0116 13:04:55.139844 25452 net.cpp:200] cifar does not need backward computation.
I0116 13:04:55.139858 25452 net.cpp:242] This network produces output accuracy
I0116 13:04:55.139873 25452 net.cpp:242] This network produces output loss
I0116 13:04:55.139910 25452 net.cpp:255] Network initialization done.
I0116 13:04:55.140070 25452 solver.cpp:56] Solver scaffolding done.
I0116 13:04:55.141544 25452 caffe.cpp:248] Starting Optimization
I0116 13:04:55.141582 25452 solver.cpp:272] Solving CIFAR10_quick
I0116 13:04:55.141597 25452 solver.cpp:273] Learning Rate Policy: fixed
I0116 13:04:55.146293 25452 solver.cpp:330] Iteration 0, Testing net (#0)
I0116 13:04:56.560989 25456 data_layer.cpp:73] Restarting data prefetching from start.
I0116 13:04:56.592571 25452 solver.cpp:397]     Test net output #0: accuracy = 0.0981
I0116 13:04:56.592634 25452 solver.cpp:397]     Test net output #1: loss = 2.303 (* 1 = 2.303 loss)


配置CIFAR10出现的错误
1、在配置cifar10 的时候，出现下列错误。
gwp@gwp:~/caffe/examples/cifar10$ ./create_cifar10.sh 
Creating lmdb...
./create_cifar10.sh: 13: ./create_cifar10.sh: ./build/examples/cifar10/convert_cifar_data.bin: not found
解决方法：
gwp@gwp:~/caffe$ ./examples/cifar10/create_cifar10.sh 
提出问题：在根目录下直接运行跟在当前目录下直接运行有什么区别？
如果还有其他类似的问题，也可以用这种方法解决，比如在mnist运行的时候。 

 
   
 
   
 
    
 
   
 
   
 
    
 
     
 
      Error：Check failed: data_file Unable to open train file #1 
         
     
 
     
 
      
  
     
 
     
 
      
  
     
 
    
 
    

      
      


      
      
windows下跑caffe的convert_cifar_data.exe进行格式转化，出错：Check failed: data_file Unable to open train file #1
      
      

       
       其实这是一个非常非常简单的错误。
      
      
 
    
 
   
 
   
 
    
 
   
 
   
 
    
 
   
 
   
 
     
    
 
   
 
   
 
    
 
   
 
   
 
    

      
      
      
      

       
       得到 cifar-10-binary.gz 
      
      

      
      

       
       错误出现的原因就在解压这个压缩包上。
      
      

      
      

       
       原因：
      
      
cifar-10-binary.gz 　是在linux下进行压缩的，windows下解压，一般实用rar，解压后得到 cifar-10-binary 是错误的，应该用7-zip进行解压，解压后得到下面的文件就对了。然后运行 convert_cifar_data.exe input output 成功。






 
   
 
   
 
    
 
   
 
   
 
    
 
   
 
   
 
    
 
   
 
   
 
    
 
   
 
  
 
  

     
     

      
      


      
      

     
     




 
 
 


                

        

        

    

  


    

      

        

            

              
                
                  苏源流
                
              
            

            

                关注
              
            

        

        

          
      

      









                



	

		

			

				
					
机器学习笔记 - 机器学习/深度学习实战案例合集

				
			

			

				

					学以致用 知行合一
				

				

					06-05
					
					1730
					
				

			

		

		

			
				
案例1、使用opencv的DNN模块调用caffe的GoogLeNet图像分类模型
案例2、使用Opencv的DNN模块调用Tensorflow的mobilenet对象检测模型
案例3、基于tensorflow2.0的手写数字识别，并导出pb模型供OpenCV的C++版本的DNN模块调用
案例4、基于OpenCV图像处理和手写数字识别进行数独求解
案例5、自定义数据集 + 标准前馈神经网络分类进行训练
案例6、自定义数据集 + 卷积神经网络分类进行训练
案例7、使用Opencv的DNN模块调用Caffe的人

			
		

	



                

              
                



	

		

			

				
					
深度学习讲座笔记：Deep Learning for Computer Vision - Andrej Karpathy at Bay Area Deep Learning School

				
			

			

				

					LL的博客
				

				

					09-26
					
					2409
					
				

			

		

		

			
				
Andrej Karpathy这次演讲是Day 1 的第2个演讲，题为深度学习在图像处理方面的应用，阐述了卷积神经网络架构的设计，以及ILSVR历年竞赛的情况以及最近在图像处理这方面的进展，并且给出了一些如何开展这方面研究工作的实用建议。

			
		

	



                


  
              

                


	

		

			

				
					
caffe学习笔记（五）：运行cifar实例

				
			

			

				

					小熊猫的博客
				

				

					07-25
					
					438
					
				

			

		

		

			
				
cifar10数据训练样本50000张，测试样本10000张，每张为32*32的彩色三通道图片，共分为10类。

下载数据：


sudo sh data/cifar10/get_cifar10.sh

运行成功后，会在 data/cifar10/文件夹下生成一堆bin文件

转换数据格式为lmdb：


sudo sh examples/cifar10/create_cifar10.sh

转换...

			
		

	





	

		

			

				
					
《深度学习——实战caffe》——运行caffe自带的Cifar10实例教程

				
			

			

				

					IT人生
				

				

					04-27
					
					343
					
				

			

		

		

			
				
CIFAR-10数据集含有6万个32*32的彩色图像，共分为10种类型，。包含50000张训练图片，10000张测试图片，数据集的数据存在一个10000*3072的numpy数组中，单位是uint8s，3072是存储了一个32*32的彩色图像。（3072=1024*3）。前1024位是r值，中间1024是g值，后面1024是b值。主要作用是小图片分类。第一步：打开caffe根目录cd ~/caff...

			
		

	



		

			
		



	

		

			

				
					
caffe 05 win10 运行examples的cifar10测试用例

				
			

			

				

					一朵花开的时间
				

				

					03-12
					
					1062
					
				

			

		

		

			
				
01 资料caffe官网Examples中有关于cifar10测试的详细介绍。 
http://caffe.berkeleyvision.org/gathered/examples/cifar10.html 
caffe github下载的代码中有关于cifar10的网络定义示例和下载数据sh文件。
  目录
  文件类型
  文件作用  examples\cifar10
  *.prototxt

			
		

	





	

		

			

				
					
【超详细】利用CIFAR10数据集 进行完整的网络模型训练

				
			

			

				

					m0_51816252的博客
				

				

					06-06
					
					4024
					
				

			

		

		

			
				
本文利用CIFAR10模型来进行进行完整的网络模型训练，可以让我们知道进行数据训练的基本过程。除此之外，本文来提到一些完整的网络模型训练的时候，应该注意的细节问题。

			
		

	





	

		

			

				
					
cifar10官方例程学习记录

				
			

			

				

					wang_kmin的博客
				

				

					08-15
					
					5265
					
				

			

		

		

			
				
整理学习到的东西，以备以后可以查看。

第一部分：训练

一、使用FLAGS设置参数

初始化FLAGS：tf.app.flags.FLAGS。

定义参数：tf.app.flags.DEFINE_xxx，xxx为参数类型。传入的第一个参数为变量名，如train_dir，可通过FLAGS.train_dir取得该变量的值。第二个参数为默认值，第三个参数为说明内容。当不设置该变量的值时，通过FLAG...

			
		

	





	

		

			

				
					
学习笔记1_CIFAR-10在caffe上进行训练与学习-薛开宇1

				
			

			

				

					08-03
				

			

		

		

			
				
这篇学习笔记主要介绍了如何在深度学习框架Caffe上使用CIFAR-10数据集进行训练和学习。CIFAR-10是一个广泛使用的图像识别数据集，包含60000张32x32彩色图像，分为10个类别，其中50000张用于训练，10000张用于测试。 ...

			
		

	





	

		

			

				
					
深度学习_21天实战Caffe.pdf

				
			

			

				

					chenxuanhanhao的专栏
				

				

					08-14
					
					2896
					
				

			

		

		

			
				
深度学习_21天实战Caffe.pdf



原

	深度学习21天实战caffe学习笔记《1:深度学习的过往》

	1. 深度学习DL： 1.1、有监督学习、无监督学习、过拟合、训练样本、泛化、训练集、验证集、测试集这些和深度学习有关的知识需要实现明白，这里有一个深度学习的资料你们可以去下载； http://download.csdn.net/detail/julialove102123...

			
		

	





	

		

			

				
					
caffe学习笔记完整版

				
			

			

				

					11-08
				

			

		

		

			
				
Caffe是一个深度学习框架，主要关注速度和模块化，尤其擅长于卷积神经网络（CNN）的研究和应用开发。以下详细知识点将围绕如何在Caffe框架上进行深度学习模型的训练、测试以及可视化等操作展开。  1. CIFAR-10 数据...

			
		

	





	

		

			

				
					
cifar10数据集下载及图片格式解析

					
热门推荐

				
			

			

				

					Briwisdom的博客
				

				

					05-15
					
					1万+
					
				

			

		

		

			
				
CIFAR-10 是由 Hinton 的学生 Alex Krizhevsky 和 Ilya Sutskever 整理的一个用于识别普适物体的小型数据集。一共包含 10 个类别的 RGB 彩色图 片：飞机（ a叩lane ）、汽车（ automobile ）、鸟类（ bird ）、猫（ cat ）、鹿（ deer ）、狗（ dog ）、蛙类（ frog ）、马（ horse ）、船（ ship ）和卡车（ truck ）。

图片的尺寸为 32×32×3 ，数据集中一共有 50...

			
		

	





	

		

			

				
					
create_cifar10.sh: 13: create_cifar10.sh: ./build/examples/cifar10/convert_cifar_data.bin: not found

				
			

			

				

					Feeryman_Lee的博客
				

				

					03-09
					
					610
					
				

			

		

		

			
				
成功解决create_cifar10.sh: 13: create_cifar10.sh: ./build/examples/cifar10/convert_cifar_data.bin: not found

			
		

	





	

		

			

				
					
学习笔记：Caffe上配置和运行Cifar10的示例

				
			

			

				

					zhazhiqiang2010的专栏
				

				

					07-28
					
					1550
					
				

			

		

		

			
				
【原文：http://blog.csdn.net/lynnandwei/article/details/44222389】


CIFAR-10数据集含有6万个32*32的彩色图像，共分为10种类型，由 Alex Krizhevsky, Vinod Nair和 Geoffrey Hinton收集而来。包含50000张训练图片，10000张测试图片

http://www.cs.toron

			
		

	





	

		

			

				
					
caffe典型识别示例CIFAR_10的运行详解

				
			

			

				

					执笔天涯
				

				

					07-03
					
					2851
					
				

			

		

		

			
				
文件说明：
1. caffe的运行示例 Cifar10
2. 分析Cifar10的数据集合模型结构
3.给出Ciffar10的运行步骤


运行环境：
Windows7 X86 + caffe + VS2013


参考资料:
1. http://blog.csdn.net/maweifei/article/details/52981425 
2. http://www.c

			
		

	





	

		

			

				
					
Ubuntu14.04LTS下应用Caffe训练Cifar10错误案例分析与解决方案1

				
			

			

				

					tricoffee
				

				

					03-21
					
					4326
					
				

			

		

		

			
				
（1）打开终端，应用cd切换路径，如 cd ~/caffe/data/cifar10
（2）继续执行命令   ./get_cifar10.sh 
（3）成功下载数据集之后，执行ls即可见所下载的数据文件，如
batches.meta.txt  data_batch_3.bin  get_cifar10.sh
data_batch_1.bin  data_batch_4.bin  readm

			
		

	





	

		

			

				
					
运行caffe自带的Cifar10实例教程

				
			

			

				

					FB0000的博客
				

				

					09-03
					
					4390
					
				

			

		

		

			
				
CIFAR-10数据集含有6万个32*32的彩色图像，共分为10种类型，。包含50000张训练图片，10000张测试图片，数据集的数据存在一个10000*3072的numpy数组中，单位是uint8s，3072是存储了一个32*32的彩色图像。（3072=1024*3）。前1024位是r值，中间1024是g值，后面1024是b值。主要作用是小图片分类。

			
		

	





	

		

			

				
					
基于ubuntu对于CIFRA-10进行训练学习

				
			

			

				

					mdjxy63的博客
				

				

					07-13
					
					283
					
				

			

		

		

			
				
参考博客：http://blog.csdn.net/c406495762/article/details/67650370




1.深度学习中经常看到epoch、 iteration和batchsize，下面按自己的理解说说这三个的区别：

    (1)batchsize：批大小。在深度学习中，一般采用SGD训练，即每次训练在训练集中取batchsize个样本训练；

			
		

	





	

		

			

				
					
【C++ Caffe】ubuntu下MNIST训练结果

				
			

			

				

					Feeryman_Lee的博客
				

				

					02-26
					
					816
					
				

			

		

		

			
				
lichunlin@ThinkPad-T420:~/caffe/data/mnist$ ./get_mnist.sh 
Downloading...
--2020-02-26 15:29:13--  http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz
Resolving yann.lecun.com (yann.lecun.com...

			
		

	





	

		

			

				
					
CIFAR-10训练教程：使用Caffe进行深度学习

				
			

			

				

					
				

			

		

		

			
				
"这份学习笔记主要介绍了如何使用深度学习工具Caffe进行训练和学习，特别是针对CIFAR-10数据集的操作步骤。CIFAR-10是一个包含60000个32x32彩色图像的数据库，分为10个类别，用于训练和测试。在Caffe中进行训练前，...

			
		

	



              




          




        



    





    



      

      

        
      

      





	

		评论	
	

	
	




  

    

      添加红包
      
    

    

      

        
        

          
          
        

        
请填写红包祝福语或标题

      

      

        
        

          
          个
        

        
红包个数最小为10个

      

      

        
        

          
          元
        

        
红包金额最低5元

      

      

        
        

          当前余额3.43元
          前往充值 >
        

      

      

        

          需支付：10.00元
        

        
        
      

    

  





  

    
    
  

  

    
    
  








  

    

      

        

          

            
            
成就一亿技术人!

          

          

        

        

          

          

            领取后你会自动成为博主和红包主的粉丝
            规则
          

        

      

      

        

          

            
              
            
            

              
hope_wisdom
 发出的红包
            

          

        

        

          

          

          

        

      

    

    

  



      
      

      
实付元

      
使用余额支付

      

        

          

            
            点击重新获取
          

        

        
扫码支付

      

      

        
          
            
          
          
        
      

      

        
        钱包余额
          0
          

            
            

              

                
抵扣说明：

                
 1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
 2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。

              

            

          

      

      余额充值