错误: ‘ImageDataGenerator’ 对象没有 ‘shape’ 属性。

huangapple
go评论98阅读模式
英文:

Error: 'ImageDataGenerator' object has no attribute 'shape'

问题

  1. 我是编程新手，一直在尝试编写用于图像分类的神经网络代码，但不幸的是，我遇到了这个错误：
  3. ```python
  4. import tensorflow as tf
  5. import tensorflow.keras as kr
  6. import os
  8. train_PATH='./Data/Entrenamiento/'
  9. valid_PATH='./Data/Validacion/'
  11. # 参数
  13. epocas=20
  14. alt,long=100,100
  15. batch_size=32
  16. pasos=1000
  17. pasos_valid=200
  18. n_filtroConv1=32
  19. n_filtroConv2=64
  20. 大小_filtro1=(3,3)
  21. 大小_filtro2=(2,2)
  22. 大小_pooling=(2,2)
  23. clases=3
  24. lr=0.0005
  26. # 图像预处理
  28. train_generador=kr.preprocessing.image.ImageDataGenerator(
  29.  rescale=1./255,
  30.  shear_range=0.3,
  31.  zoom_range=0.3,
  32.  horizontal_flip=True)
  34. valid_generador=kr.preprocessing.image.ImageDataGenerator(
  35.  rescale=1./255)
  37. imagen_entrenamiento=train_generador.flow_from_directory(
  38.  train_PATH,
  39.  target_size=(alt,long),
  40.  batch_size=batch_size,
  41.  class_mode='categorical')
  43. imagen_validacion=valid_generador.flow_from_directory(
  44.  valid_PATH,
  45.  target_size=(alt,long),
  46.  batch_size=batch_size,
  47.  class_mode='categorical')
  49. # 创建卷积神经网络
  51. nn=kr.Sequential()
  53. nn.add(kr.layers.Convolution2D(n_filtroConv1,大小_filtro1,padding='same',input_shape=(alt,long,3),activation='relu'))
  54. nn.add(kr.layers.MaxPooling2D(pool_size=大小_pooling))
  55. nn.add(kr.layers.Convolution2D(n_filtroConv2,大小_filtro2,padding='same',activation='relu'))
  56. nn.add(kr.layers.MaxPooling2D(pool_size=大小_pooling))
  58. nn.add(kr.layers.Flatten())
  59. nn.add(kr.layers.Dense(256,activation='relu'))
  60. nn.add(kr.layers.Dropout(0.5))
  61. nn.add(kr.layers.Dense(clases,activation='softmax'))
  63. nn.compile(loss='categorical_crossentropy',optimizer=tf.keras.optimizers.Adam(lr=lr),metrics=['accuracy'])
  65. nn.fit_generator(train_generador,steps_per_epoch=pasos,epochs=epocas,validation_data=imagen_validacion,validation_steps=pasos_valid)
  67. dir_modelo='./modelo/'
  69. if not os.path.exists(dir_modelo):
  70.  os.mkdir(dir_modelo)
  72. nn.save(dir_modelo+'modelo.h5')
  73. nn.save_weights(dir_modelo+'pesos.h5')
  75. 当我在Anaconda-Navigator上的Python 3.7中使用Spyder执行此代码时，我遇到了这个错误：
  77. ```python
  78. nn.fit_generator(train_generador,steps_per_epoch=pasos,epochs=epocas,validation_data=imagen_validacion,validation_steps=pasos_valid)
  80.  File "/Applications/anaconda3/lib/python3.7/site-packages/tensorflow_core/python/keras/engine/training.py", line 1297, in fit_generator
  81. steps_name='steps_per_epoch')
  83.  File "/Applications/anaconda3/lib/python3.7/site-packages/tensorflow_core/python/keras/engine/training_generator.py", line 144, in model_iteration
  84. shuffle=shuffle)
  86.  File "/Applications/anaconda3/lib/python3.7/site-packages/tensorflow_core/python/keras/engine/training_generator.py", line 477, in convert_to_generator_like
  87. num_samples = int(nest.flatten(data)[0].shape[0])
  89. AttributeError: 'ImageDataGenerator' object has no attribute 'shape'
英文:

I am new in coding and I've been trying to code this neural network for images classification, but unfortunately, I've encountered this error:

  1. import tensorflow as tf
  2. import tensorflow.keras as kr
  3. import os
  6. train_PATH='./Data/Entrenamiento/'
  7. valid_PATH='./Data/Validacion/'
  9.  # Parámetros
  11. epocas=20
  12. alt,long=100,100
  13. batch_size=32
  14. pasos=1000
  15. pasos_valid=200
  16. n_filtroConv1=32
  17. n_filtroConv2=64
  18. tamaño_filtro1=(3,3)
  19. tamaño_filtro2=(2,2)
  20. tamaño_pooling=(2,2)
  21. clases=3
  22. lr=0.0005
  24.  # Pre-procesamiento de imágenes
  26. train_generador=kr.preprocessing.image.ImageDataGenerator(
  27.  rescale=1./255,
  28.  shear_range=0.3,
  29.  zoom_range=0.3,
  30.  horizontal_flip=True)
  32. valid_generador=kr.preprocessing.image.ImageDataGenerator(
  33.  rescale=1./255)
  35. imagen_entrenamiento=train_generador.flow_from_directory(
  36.  train_PATH,
  37.  target_size=(alt,long),
  38.  batch_size=batch_size,
  39.  class_mode='categorical')
  41. imagen_validacion=valid_generador.flow_from_directory(
  42.  valid_PATH,
  43.  target_size=(alt,long),
  44.  batch_size=batch_size,
  45.  class_mode='categorical')
  47.  # Crear red convolucional
  49. nn=kr.Sequential()
  51. nn.add(kr.layers.Convolution2D(n_filtroConv1,tamaño_filtro1,padding='same',input_shape=(alt,long,3),activation='relu'))
  52. nn.add(kr.layers.MaxPooling2D(pool_size=tamaño_pooling))
  53. nn.add(kr.layers.Convolution2D(n_filtroConv2,tamaño_filtro2,padding='same',activation='relu'))
  54. nn.add(kr.layers.MaxPooling2D(pool_size=tamaño_pooling))
  56. nn.add(kr.layers.Flatten())
  57. nn.add(kr.layers.Dense(256,activation='relu'))
  58. nn.add(kr.layers.Dropout(0.5))
  59. nn.add(kr.layers.Dense(clases,activation='softmax'))
  61. nn.compile(loss='categorical_crossentropy',optimizer=tf.keras.optimizers.Adam(lr=lr),metrics=['accuracy'])
  63. nn.fit_generator(train_generador,steps_per_epoch=pasos,epochs=epocas,validation_data=imagen_validacion,validation_steps=pasos_valid)
  65. dir_modelo='./modelo/'
  67. if not os.path.exists(dir_modelo):
  68.  os.mkdir(dir_modelo)
  70. nn.save(dir_modelo+'modelo.h5')
  71. nn.save_weights(dir_modelo+'pesos.h5')

When I execute this code on Spyder with Python 3.7 on Anaconda-Navigator, I'm facing with this error:

  1. nn.fit_generator(train_generador,steps_per_epoch=pasos,epochs=epocas,validation_data=imagen_validacion,validation_steps=pasos_valid)
  3.  File "/Applications/anaconda3/lib/python3.7/site-packages/tensorflow_core/python/keras/engine/training.py", line 1297, in fit_generator
  4. steps_name='steps_per_epoch')
  6.  File "/Applications/anaconda3/lib/python3.7/site-packages/tensorflow_core/python/keras/engine/training_generator.py", line 144, in model_iteration
  7. shuffle=shuffle)
  9.  File "/Applications/anaconda3/lib/python3.7/site-packages/tensorflow_core/python/keras/engine/training_generator.py", line 477, in convert_to_generator_like
  10. num_samples = int(nest.flatten(data)[0].shape[0])
  12. AttributeError: 'ImageDataGenerator' object has no attribute 'shape'

答案1

得分: 1

生成器是 imagen_entrenamiento，而不是 train_generator。

使用 nn.fit_generator(imagen_entrenamiento,...)。

英文:

The generator is imagen_entrenamiento, not train_generator.

Use nn.fit_generator(imagen_entrenamiento,...)

huangapple
  • 本文由 发表于 2020年1月6日 21:44:38
  • 转载请务必保留本文链接：https://go.coder-hub.com/59613222.html
  • keras
  • python
  • tensorflow
  • tensorflow2.0
匿名

发表评论

匿名网友

:?: :razz: :sad: :evil: :!: :smile: :oops: :grin: :eek: :shock: :???: :cool: :lol: :mad: :twisted: :roll: :wink: :idea: :arrow: :neutral: :cry: :mrgreen:

确定