DL之CNN：利用卷积神经网络算法(2→2,基于Keras的API-Functional)利用MNIST(手写数字图片识别)数据集实现多分类预测

乔乔乔
乔乔乔 2022-09-19 14:16:11 53099
分类专栏： 资讯

DL之CNN：利用卷积神经网络算法(2→2,基于Keras的API-Functional)利用MNIST(手写数字图片识别)数据集实现多分类预测

目录

输出结果

设计思路

核心代码

输出结果

下边两张图对应查看，可知，数字0有965个是被准确识别到！

  1. 1.10.0
  2. Size of:
  3. - Training-set:		55000
  4. - Validation-set:	5000
  5. - Test-set:		10000
  6. Epoch 1/1
  7.  
  8.   128/55000 [..............................] - ETA: 14:24 - loss: 2.3439 - acc: 0.0938
  9.   256/55000 [..............................] - ETA: 14:05 - loss: 2.2695 - acc: 0.1016
  10.   384/55000 [..............................] - ETA: 13:20 - loss: 2.2176 - acc: 0.1302
  11.   512/55000 [..............................] - ETA: 13:30 - loss: 2.1608 - acc: 0.2109
  12.   640/55000 [..............................] - ETA: 13:29 - loss: 2.0849 - acc: 0.2500
  13.   768/55000 [..............................] - ETA: 13:23 - loss: 2.0309 - acc: 0.2734
  14.   896/55000 [..............................] - ETA: 13:30 - loss: 1.9793 - acc: 0.2946
  15.  1024/55000 [..............................] - ETA: 13:23 - loss: 1.9105 - acc: 0.3369
  16.  1152/55000 [..............................] - ETA: 13:22 - loss: 1.8257 - acc: 0.3776
  17. ……
  18. 53760/55000 [============================>.] - ETA: 18s - loss: 0.2106 - acc: 0.9329
  19. 53888/55000 [============================>.] - ETA: 16s - loss: 0.2103 - acc: 0.9330
  20. 54016/55000 [============================>.] - ETA: 14s - loss: 0.2100 - acc: 0.9331
  21. 54144/55000 [============================>.] - ETA: 13s - loss: 0.2096 - acc: 0.9333
  22. 54272/55000 [============================>.] - ETA: 11s - loss: 0.2092 - acc: 0.9334
  23. 54400/55000 [============================>.] - ETA: 9s - loss: 0.2089 - acc: 0.9335 
  24. 54528/55000 [============================>.] - ETA: 7s - loss: 0.2086 - acc: 0.9336
  25. 54656/55000 [============================>.] - ETA: 5s - loss: 0.2082 - acc: 0.9337
  26. 54784/55000 [============================>.] - ETA: 3s - loss: 0.2083 - acc: 0.9337
  27. 54912/55000 [============================>.] - ETA: 1s - loss: 0.2082 - acc: 0.9337
  28. 55000/55000 [==============================] - 837s 15ms/step - loss: 0.2080 - acc: 0.9338
  29.  
  30.    32/10000 [..............................] - ETA: 21s
  31.   160/10000 [..............................] - ETA: 8s 
  32.   288/10000 [..............................] - ETA: 6s
  33.   448/10000 [>.............................] - ETA: 5s
  34.   576/10000 [>.............................] - ETA: 5s
  35.   736/10000 [=>............................] - ETA: 4s
  36.   864/10000 [=>............................] - ETA: 4s
  37.  1024/10000 [==>...........................] - ETA: 4s
  38.  1152/10000 [==>...........................] - ETA: 4s
  39.  1312/10000 [==>...........................] - ETA: 4s
  40.  1440/10000 [===>..........................] - ETA: 4s
  41.  1600/10000 [===>..........................] - ETA: 3s
  42.  1728/10000 [====>.........................] - ETA: 3s
  43. ……
  44.  3008/10000 [========>.....................] - ETA: 3s
  45.  3168/10000 [========>.....................] - ETA: 3s
  46.  3296/10000 [========>.....................] - ETA: 3s
  47.  3456/10000 [=========>....................] - ETA: 2s
  48. ……
  49.  5248/10000 [==============>...............] - ETA: 2s
  50.  5376/10000 [===============>..............] - ETA: 2s
  51.  5536/10000 [===============>..............] - ETA: 2s
  52.  5664/10000 [===============>..............] - ETA: 1s
  53.  5792/10000 [================>.............] - ETA: 1s
  54. ……
  55.  7360/10000 [=====================>........] - ETA: 1s
  56.  7488/10000 [=====================>........] - ETA: 1s
  57.  7648/10000 [=====================>........] - ETA: 1s
  58.  7776/10000 [======================>.......] - ETA: 1s
  59.  7936/10000 [======================>.......] - ETA: 0s
  60.  8064/10000 [=======================>......] - ETA: 0s
  61.  8224/10000 [=======================>......] - ETA: 0s
  62. ……
  63.  9760/10000 [============================>.] - ETA: 0s
  64.  9920/10000 [============================>.] - ETA: 0s
  65. 10000/10000 [==============================] - 4s 449us/step
  66. loss 0.05686537345089018
  67. acc 0.982
  68. acc: 98.20%
  69. [[ 965    0    4    0    0    0    4    1    2    4]
  70.  [   0 1128    3    0    0    0    0    1    3    0]
  71.  [   0    0 1028    0    0    0    0    1    3    0]
  72.  [   0    0   10  991    0    2    0    2    3    2]
  73.  [   0    0    3    0  967    0    1    1    1    9]
  74.  [   2    0    1    7    1  863    5    1    4    8]
  75.  [   2    3    0    0    3    2  946    0    2    0]
  76.  [   0    1   17    1    1    0    0  987    2   19]
  77.  [   2    0    9    2    0    1    0    1  955    4]
  78.  [   1    4    3    2    8    0    0    0    1  990]]
  79.  
  80.  
  81. _________________________________________________________________
  82. Layer (type)                 Output Shape              Param    
  83. =================================================================
  84. input_1 (InputLayer)         (None, 784)               0         
  85. _________________________________________________________________
  86. reshape (Reshape)            (None, 28, 28, 1)         0         
  87. _________________________________________________________________
  88. layer_conv1 (Conv2D)         (None, 28, 28, 16)        416       
  89. _________________________________________________________________
  90. max_pooling2d (MaxPooling2D) (None, 14, 14, 16)        0         
  91. _________________________________________________________________
  92. layer_conv2 (Conv2D)         (None, 14, 14, 36)        14436     
  93. _________________________________________________________________
  94. max_pooling2d_1 (MaxPooling2 (None, 7, 7, 36)          0         
  95. _________________________________________________________________
  96. flatten (Flatten)            (None, 1764)              0         
  97. _________________________________________________________________
  98. dense (Dense)                (None, 128)               225920    
  99. _________________________________________________________________
  100. dense_1 (Dense)              (None, 10)                1290      
  101. =================================================================
  102. Total params: 242,062
  103. Trainable params: 242,062
  104. Non-trainable params: 0
  105. _________________________________________________________________
  106. (5, 5, 1, 16)
  107. (1, 28, 28, 16)

设计思路

核心代码

后期更新……

  1. path_model = 'Functional_model.keras'  
  2.                   
  3. from tensorflow.python.keras.models import load_model  
  4. model2_1 = load_model(path_model)      
  5.  
  6. model_weights_path = 'Functional_model_weights.keras'
  7. model2_1.save_weights(model_weights_path )                  
  8. model2_1.load_weights(model_weights_path, by_name=True ) 
  9. model2_1.load_weights(model_weights_path)  
  10.  
  11.  
  12. result = model.evaluate(x=data.x_test,
  13.                         y=data.y_test)
  14.   
  15. for name, value in zip(model.metrics_names, result):
  16.     print(name, value)
  17. print("{0}: {1:.2%}".format(model.metrics_names[1], result[1]))
  18.  
  19.  
  20. y_pred = model.predict(x=data.x_test) 
  21. cls_pred = np.argmax(y_pred, axis=1)   
  22. plot_example_errors(cls_pred)        
  23. plot_confusion_matrix(cls_pred)     
  24.  
  25.  
  26.  
  27. images = data.x_test[0:9]                      
  28. cls_true = data.y_test_cls[0:9]                 
  29. y_pred = model.predict(x=images)               
  30. cls_pred = np.argmax(y_pred, axis=1)            
  31. title = 'MNIST(Sequential Model): plot predicted example, resl VS predict'
  32. plot_images(title, images=images,               
  33.             cls_true=cls_true,
  34.             cls_pred=cls_pred)

网站声明：如果转载，请联系本站管理员。否则一切后果自行承担。

本文链接：https://www.xckfsq.com/news/show.html?id=2917
赞同 0
评论 0 条

相关文章

乔乔乔
乔乔乔L0
粉丝 0 发表 9 + 关注 私信
上周热门
如何使用 StarRocks 管理和优化数据湖中的数据？  2935
【软件正版化】软件正版化工作要点  2854
统信UOS试玩黑神话：悟空  2811
信刻光盘安全隔离与信息交换系统  2702
镜舟科技与中启乘数科技达成战略合作，共筑数据服务新生态  1235
grub引导程序无法找到指定设备和分区  1205
点击报名 | 京东2025校招进校行程预告  162
华为全联接大会2024丨软通动力分论坛精彩议程抢先看！  160
2024海洋能源产业融合发展论坛暨博览会同期活动-海洋能源与数字化智能化论坛成功举办  156
华为纯血鸿蒙正式版9月底见！但Mate 70的内情还得接着挖...  154
本周热议
我的信创开放社区兼职赚钱历程 40
今天你签到了吗？ 27
信创开放社区邀请他人注册的具体步骤如下 15
如何玩转信创开放社区—从小白进阶到专家 15
方德桌面操作系统 14
我有15积分有什么用? 13
用抖音玩法闯信创开放社区——用平台宣传企业产品服务 13
如何让你先人一步获得悬赏问题信息？（创作者必看） 12
2024中国信创产业发展大会暨中国信息科技创新与应用博览会 9
中央国家机关政府采购中心：应当将CPU、操作系统符合安全可靠测评要求纳入采购需求 8
热门标签更多

加入交流群

请使用微信扫一扫!