关于机器学习中的时代和图像 [重复]

Question

这个问题在这里已经有了答案：

Keras 未对整个数据集进行训练3 个回答

10 个月前关闭。

当我传递给 CNN 模型时，我在 train_images 中有 186 张图像，在 valid_images 中有 174 张图像 它只训练 6 张图像。 我没有创建任何批量大小。 数据集名称是Lego minifigure 。

 '''
    print(train_images.shape)
    print(type(train_images))
    print(valid_images.shape)
    print(type(valid_images))
    print(train_targets.shape)
    print(valid_targets.shape)
    print(type(train_targets))
    print(type(valid_targets))
    '''
    output is 
    
    (186, 20, 20, 3)
    <class 'numpy.ndarray'>
    (174, 20, 20, 3)
    <class 'numpy.ndarray'>
    (186, 33)
    (174, 33)
    <class 'numpy.ndarray'>
    <class 'numpy.ndarray'>
    '''
    model
    model=tf.keras.Sequential()
    model.add(tf.keras.layers.Conv2D(20,(3,3),activation='relu',input_shape=(20,20,3)))
    model.add(tf.keras.layers.MaxPooling2D(2,2))
    model.add(tf.keras.layers.Flatten())
    model.add(tf.keras.layers.Dense(100,activation='relu'))
    model.add(tf.keras.layers.Dense(33,activation='softmax'))
    model.compile(loss='categorical_crossentropy',metrics=['accuracy'],optimizer='adam')
    model.summary()
    '''
    
    Model: "sequential_3"
    _________________________________________________________________
    Layer (type)                 Output Shape              Param #   
    =================================================================
    conv2d_3 (Conv2D)            (None, 18, 18, 20)        560       
    _________________________________________________________________
    max_pooling2d_3 (MaxPooling2 (None, 9, 9, 20)          0         
    _________________________________________________________________
    flatten_3 (Flatten)          (None, 1620)              0         
    _________________________________________________________________
    dense_4 (Dense)              (None, 100)               162100    
    _________________________________________________________________
    dense_5 (Dense)              (None, 33)                3333      
    =================================================================
    Total params: 165,993
    Trainable params: 165,993
    Non-trainable params: 0
    ___________________________
    '''
    hist=model.fit(train_images,train_targets,epochs=100,validation_data=(valid_images,valid_targets))
    
    '''
    Epoch 1/100
    6/6 [==============================] - 0s 10ms/step - loss: 2.7642 - accuracy: 0.4355 - val_loss: 3.1673 - val_accuracy: 0.1839

为什么它只训练 6 张图像？ 我是 Ml 的初学者，所以如果有人帮助，那将是很大的帮助！

Answer 1

默认情况下，batch_size 等于 32