【神經網絡與深度學習】Caffe部署中的幾個train-test-solver-prototxt-deploy等說明

1:神經網絡中,咱們經過最小化神經網絡來訓練網絡,因此在訓練時最後一層是損失函數層(LOSS),css

在測試時咱們經過準確率來評價該網絡的優劣,所以最後一層是準確率層(ACCURACY)。html

可是當咱們真正要使用訓練好的數據時,咱們須要的是網絡給咱們輸入結果,對於分類問題,咱們須要得到分類結果,以下右圖最後一層咱們獲得python

的是機率,咱們不須要訓練及測試階段的LOSS,ACCURACY層了。網絡

下圖是能過$CAFFE_ROOT/python/draw_net.py繪製$CAFFE_ROOT/models/caffe_reference_caffnet/train_val.prototxt   , $CAFFE_ROOT/models/caffe_reference_caffnet/deploy.prototxt,分別表明訓練時與最後使用時的網絡結構。ide

 

咱們通常將train與test放在同一個.prototxt中,須要在data層輸入數據的source,函數

而在使用時.prototxt只須要定義輸入圖片的大小通道數據參數便可,以下圖所示,分別是學習

$CAFFE_ROOT/models/caffe_reference_caffnet/train_val.prototxt   , $CAFFE_ROOT/models/caffe_reference_caffnet/deploy.prototxt的data層測試

訓練時, solver.prototxt中使用的是rain_val.prototxtui

1
./build/tools/caffe/train -solver ./models/bvlc_reference_caffenet/solver.prototxt

 使用上面訓練的網絡提取特徵,使用的網絡模型是deploy.prototxtgoogle

1
./build/tools/extract_features.bin models/bvlc_refrence_caffenet.caffemodel models/bvlc_refrence_caffenet/deploy.prototxt

      。。

 

2:

(1)介紹 *_train_test.prototxt文件與 *_deploy.prototxt文件的不http://blog.csdn.net/sunshine_in_moon/article/details/49472901    

(2)生成deploy文件的Python代碼:http://www.cnblogs.com/denny402/p/5685818.html       

*_train_test.prototxt文件:這是訓練與測試網絡配置文件

 
*_deploy.prototxt文件:這是模型構造文件
 

在博文http://www.cnblogs.com/denny402/p/5685818.html     中給出了生成 deploy.prototxt文件的Python源代碼,可是每一個網絡不一樣,修改起來比較麻煩,下面給出該博文中以mnist爲例生成deploy文件的源代碼,可根據本身網絡的設置作出相應修改:(下方代碼未測試)

# -*- coding: utf-8 -*-

from caffe import layers as L,params as P,to_proto
root='/home/xxx/'
deploy=root+'mnist/deploy.prototxt'    #文件保存路徑

def create_deploy():
    #少了第一層,data層
    conv1=L.Convolution(bottom='data', kernel_size=5, stride=1,num_output=20, pad=0,weight_filler=dict(type='xavier'))
    pool1=L.Pooling(conv1, pool=P.Pooling.MAX, kernel_size=2, stride=2)
    conv2=L.Convolution(pool1, kernel_size=5, stride=1,num_output=50, pad=0,weight_filler=dict(type='xavier'))
    pool2=L.Pooling(conv2, pool=P.Pooling.MAX, kernel_size=2, stride=2)
    fc3=L.InnerProduct(pool2, num_output=500,weight_filler=dict(type='xavier'))
    relu3=L.ReLU(fc3, in_place=True)
    fc4 = L.InnerProduct(relu3, num_output=10,weight_filler=dict(type='xavier'))
    #最後沒有accuracy層,但有一個Softmax層
    prob=L.Softmax(fc4)
    return to_proto(prob)
def write_deploy(): 
    with open(deploy, 'w') as f:
        f.write('name:"Lenet"\n')
        f.write('input:"data"\n')
        f.write('input_dim:1\n')
        f.write('input_dim:3\n')
        f.write('input_dim:28\n')
        f.write('input_dim:28\n')
        f.write(str(create_deploy()))
if __name__ == '__main__':
    write_deploy()

 

用代碼生成deploy文件仍是比較麻煩。咱們在構建深度學習網絡時,確定會先定義好訓練與測試網絡的配置文件——*_train_test.prototxt文件,咱們能夠經過修改*_train_test.prototxt文件 來生成 deploy 文件。以cifar10爲例先簡單介紹一下二者的區別。

 

(1)deploy 文件中的數據層更爲簡單,即將*_train_test.prototxt文件中的輸入訓練數據lmdb與輸入測試數據lmdb這兩層刪除,取而代之的是,

  1. layer {  
  2.   name: "data"  
  3.   type: "Input"  
  4.   top: "data"  
  5.   input_param { shape: { dim: 1 dim: 3 dim: 32 dim: 32 } }  
 
注:shape: { dim: 1 dim: 3 dim: 32 dim: 32 }表明含義:
 
shape {
  dim: 1  #num,可自行定義
  dim: 3  #通道數,表示RGB三個通道
  dim: 32   #圖像的長和寬,經過 *_train_test.prototxt文件中數據輸入層的crop_size獲取
  dim: 32
 

 

(2)卷積層和全鏈接層中weight_filler{}與bias_filler{}兩個參數不用再填寫,由於這兩個參數的值,由已經訓練好的模型*.caffemodel文件提供。以下所示代碼,將*_train_test.prototxt文件中的weight_filler、bias_filler所有刪除。

layer {  # weight_filler、bias_filler刪除  

name: "ip2"  

type: "InnerProduct"  

bottom: "ip1"   top: "ip2"  

param {    

lr_mult: 1   #權重w的學習率倍數  

}  

param {     lr_mult: 2    #偏置b的學習率倍數  

}  

inner_product_param {     num_output: 10    

weight_filler {       type: "gaussian"       std: 0.1     }    

bias_filler {       type: "constant"     }  

}

}

刪除後變爲

  1. layer {                               
  2.   name: "ip2"  
  3.   type: "InnerProduct"  
  4.   bottom: "ip1"  
  5.   top: "ip2"  
  6.   param {  
  7.     lr_mult: 1  
  8.   }  
  9.   param {  
  10.     lr_mult: 2  
  11.   }  
  12.   inner_product_param {  
  13.     num_output: 10  
  14.   }  
  15. }  
 
(3)輸出層的變化  
     1)沒有了test模塊測試精度 ,將該層刪除     
     2)輸出層
 
1)*_deploy.prototxt文件的構造和*_train_test.prototxt文件的構造最爲明顯的不一樣點是,deploy文件沒有test網絡中的test模塊,只有訓練模塊,即將*_train_test.prototxt中最後部分的test模塊測試精度刪除,即將以下代碼刪除。
 
  1. layer {                                  #刪除該層  
  2.   name: "accuracy"  
  3.   type: "Accuracy"  
  4.   bottom: "ip2"  
  5.   bottom: "label"  
  6.   top: "accuracy"  
  7.   include {  
  8.     phase: TEST  
  9.   }  
  10. }  

2) 輸出層 

*_train_test.prototxt文件

  1. layer{  
  2.   name: "loss"   #注意此處層名稱與下面的不一樣  
  3.   type: "SoftmaxWithLoss"  #注意此處與下面的不一樣  
  4.   bottom: "ip2"  
  5.   bottom: "label"    #注意標籤項在下面沒有了,由於下面的預測屬於哪一個標籤,所以不能提供標籤  
  6.   top: "loss"  
  7. }  
 
*_deploy.prototxt文件
[python]
  1. layer {  
  2.   name: "prob"  
  3.   type: "Softmax"  
  4.   bottom: "ip2"  
  5.   top: "prob"  
  6. }  

注意在兩個文件中輸出層的類型都發生了變化一個是SoftmaxWithLoss,另外一個是Softmax。另外爲了方便區分訓練與應用輸出,訓練是輸出時是loss,應用時是prob。

下面給出CIFAR10中的配置文件cifar10_quick_train_test.prototxt與其模型構造文件  cifar10_quick.prototxt 直觀展現二者的區別。

cifar10_quick_train_test.prototxt文件代碼

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
cifar10_quick_train_test.prototxt文件代碼
 
name:  "CIFAR10_quick"
  layer {               #該層去掉
   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 {             #該層去掉
   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 {                        #將下方的weight_filler、bias_filler所有刪除
   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 {                         #weight_filler、bias_filler刪除
   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 {                         #weight_filler、bias_filler刪除
   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 {                       #weight_filler、bias_filler刪除
   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 {                              # weight_filler、bias_filler刪除
   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 {                                  #將該層刪除
   name:  "accuracy"
    type:  "Accuracy"
    bottom:  "ip2"
    bottom:  "label"
    top:  "accuracy"
    include {
      phase: TEST
    }
  }
  layer {                                 #修改
   name:  "loss"        #---loss  修改成  prob
    type:  "SoftmaxWithLoss"              # SoftmaxWithLoss 修改成 softmax
    bottom:  "ip2"
    bottom:  "label"           #去掉
   top:  "loss"
  }
 
 
 
 
 
如下爲cifar10_quick.prototxt
 
layer {               #將兩個輸入層修改成該層
   name:  "data"
    type:  "Input"
    top:  "data"
    input_param { shape: { dim: 1 dim: 3 dim: 32 dim: 32 } }      #注意shape中變量值的修改,CIFAR10中的 *_train_test.protxt文件中沒有 crop_size
  }
  layer {
    name:  "conv1"
    type:  "Convolution"
    bottom:  "data"
    top:  "conv1"
    param {
      lr_mult: 1   #權重W的學習率倍數
}
    param {
      lr_mult: 2   #偏置b的學習率倍數
   }
    convolution_param {
      num_output: 32
      pad: 2   #加邊爲2
     kernel_size: 5
      stride: 1
    }
  }
  layer {
    name:  "pool1"
    type:  "Pooling"
    bottom:  "conv1"
    top:  "pool1"
    pooling_param {
      pool: MAX    #Max Pooling
     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
    }
  }
  layer {
    name:  "relu2"
    type:  "ReLU"
    bottom:  "conv2"
    top:  "conv2"
  }
  layer {
    name:  "pool2"
    type:  "Pooling"
    bottom:  "conv2"
    top:  "pool2"
    pooling_param {
      pool: AVE   #均值池化
     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
    }
  }
  layer {
    name:  "relu3"
    type:  "ReLU"   #使用ReLU激勵函數,這裏須要注意的是,本層的bottom和top都是conv3>
    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
    }
  }
  layer {
    name:  "ip2"
    type:  "InnerProduct"
    bottom:  "ip1"
    top:  "ip2"
    param {
      lr_mult: 1
    }
    param {
      lr_mult: 2
    }
    inner_product_param {
      num_output: 10
    }
  }
layer {
    name:  "prob"
    type:  "Softmax"
    bottom:  "ip2"
    top:  "prob"
  }

 3:

將train_val.prototxt 轉換成deploy.prototxt

1.刪除輸入數據(如:type:data...inckude{phase: TRAIN}),而後添加一個數據維度描述。

  1. input: "data"   
  2. input_dim: 1   
  3. input_dim: 3   
  4. input_dim: 224   
  5. input_dim: 224  
  6. force_backward: true  
2.移除最後的<span style="line-height: 24px; color: rgb(68, 68, 68); font-family: "Open Sans", Helvetica, Arial, sans-serif; font-size: 14px;">「loss」 和「accuracy」 層,加入「prob」層。</span>

 [plain]

  1. layers {  
  2.   name: "prob"  
  3.   type: SOFTMAX  
  4.   bottom: "fc8"  
  5.   top: "prob"  
  6. }  
若是train_val文件中還有其餘的預處理層,就稍微複雜點。以下,在'data'層,在‘data’層和‘conv1’層<span style="line-height: 24px; color: rgb(68, 68, 68); font-family: "Open Sans", Helvetica, Arial, sans-serif; font-size: 14px;">(with <span style="margin: 0px; padding: 0px; border: 0px currentcolor; vertical-align: baseline;">bottom:」data」  / top:」conv1″). 插入一個層來計算輸入數據的均值。</span></span>
  1. layer {  
  2. name: 「mean」  
  3. type: 「Convolution」  
  4. <strong>bottom: 「data」  
  5. top: 「data」</strong>  
  6. param {  
  7. lr_mult: 0  
  8. decay_mult: 0  
  9. }  
  10.   
  11. …}  
<span style="line-height: 1.5; margin: 0px; padding: 0px; border: 0px currentcolor; vertical-align: baseline;">在deploy.prototxt文件中,「mean」 層必須保留,只是容器改變,相應的‘conv1’也要改變<span style="line-height: 24px; color: rgb(68, 68, 68); font-family: "Open Sans", Helvetica, Arial, sans-serif; font-size: 14px;"> ( <span style="margin: 0px; padding: 0px; border: 0px currentcolor; vertical-align: baseline;"><span style="line-height: 1.5; margin: 0px; padding: 0px; border: 0px currentcolor; vertical-align: baseline;">bottom:」mean」/ <span style="line-height: 24px; margin: 0px; padding: 0px; border: 0px currentcolor; vertical-align: baseline;">top:」conv1″ )。</span></span></span></span></span>

 [plain]

  1. layer {  
  2. name: 「mean」  
  3. type: 「Convolution」  
  4. <strong>bottom: 「data」  
  5. top: 「mean「</strong>  
  6. param {  
  7. lr_mult: 0  
  8. decay_mult: 0  
  9. }  
相關文章
相關標籤/搜索