Question

我是Keras的新手，我正在尝试用Python训练人脸检测机。如您所见，生成器返回了值，但似乎输出格式不正确。任何建议都将受到赞赏

完整的ValueError如下：

ValueError：生成器的输出应为元组(x, y, sample_weight) 或(x, y)。找到：[[[[0.10196079 0.08235294 0.07058824] [0.10196079 0.08235294 0.07058824] [0.10196079 0.08235294 0.07058824] ... [0.10196079 0.08235294 0.07058824] [0.10196079 0.08235294 0.07058824] [0.10196079 0.08235294 0.07058824]

这是回溯

文件“ C：/Users/user/PycharmProjects/untitled4/transferLearning.py”，行> 103，在callbacks = [checkpoint，early]中）
    包装中的文件“ C：\ Users \ user \ Anaconda3 \ lib \ site-> packages \ keras \ legacy \ interfaces.py”，行91      return func（* args，** kwargs）
   fit_generator中的文件“ C：\ Users \ user \ Anaconda3 \ lib \ site-packages \ keras \ engine \ training.py”，行1418       initial_epoch = initial_epoch）
  在fit_generator中的文件198行中的文件“ C：\ Users \ user \ Anaconda3 \ lib \ site-> packages \ keras \ engine \ training_generator.py”      str（generator_output））

下面的完整代码

image_dir = path.join(root_dir, 'train_countinghead', 'image_data')

img_width, img_height = 256, 256
train_csv = pandas.read_csv(path.join(root_dir, 'train_countinghead', 'train.csv'))
test_csv = pandas.read_csv(path.join(root_dir, 'test_headcount.csv'))

train_samples = len(train_csv)
test_samples = len(test_csv)
batch_size = 16
epochs = 50

model = applications.VGG16(weights='imagenet', include_top=False, input_shape=(img_width, img_height, 3))

# Freeze the layers which you don't want to train. Here I am freezing the first 5 layers.
for layer in model.layers[:5]:
    layer.trainable = False

# Adding custom Layers
x = model.output
x = Flatten()(x)
x = Dense(1024, activation="relu")(x)
x = Dropout(0.5)(x)
x = Dense(1024, activation="relu")(x)
predictions = Dense(16, activation="softmax")(x)

# creating the final model
model_final = Model(inputs=model.input, outputs=predictions)

# compile the model
model_final.compile(loss="categorical_crossentropy", optimizer=optimizers.SGD(lr=0.0001, momentum=0.9),
                    metrics=["accuracy"])

# Initiate the train and test generators with data Augumentation
train_datagen = ImageDataGenerator(
    rescale=1./255,
    horizontal_flip=True,
    fill_mode="nearest",
    zoom_range=0.3,
    width_shift_range=0.3,
    height_shift_range=0.3,
    rotation_range=30
)

test_datagen = ImageDataGenerator(
    rescale=1. / 255,
    horizontal_flip=True,
    fill_mode="nearest",
    zoom_range=0.3,
    width_shift_range=0.3,
    height_shift_range=0.3,
    rotation_range=30
)

# if `class_mode` is `"categorical"` (default value) it must include the `y_col` column with the class/es of each image.
# Check the comments in method definition for more

train_generator = train_datagen.flow_from_dataframe(
    dataframe=train_csv,
    directory=image_dir,
    x_col='Name',
    target_size=(img_height, img_width),
    batch_size=batch_size,
    class_mode=None
)

test_generator = test_datagen.flow_from_dataframe(
    dataframe=test_csv,
    directory=image_dir,
    x_col='Name',
    target_size=(img_height, img_width),
    batch_size=batch_size,
    class_mode=None
)

# Save the model according to the conditions
checkpoint = ModelCheckpoint(path.join(root_dir, "vgg16_1.h5"), monitor='val_acc', verbose=1, save_best_only=True,
                             save_weights_only=False,
                             mode='auto', period=1)
early = EarlyStopping(monitor='val_acc', min_delta=0, patience=10, verbose=1, mode='auto')

# Train the model
model_final.fit_generator(
    train_generator,
    # samples_per_epoch=train_samples,
    steps_per_epoch=train_samples / batch_size,
    epochs=epochs,
    validation_data=test_generator,
    validation_steps=test_samples / batch_size,
    callbacks=[checkpoint, early])

Answer 1

问题在于，您在此处未提供目标列。如果您查看the documentation，会发现您需要（因为正在训练模型）指定y_col，并且没有class_mode=None（仅用于预测），至少对于train_generator（我不知道您打算如何使用test_generator）。

您还可能已经看到使用该错误，它告诉您并没有获取所有必需的元素（x数据，y标签）。

ValueError：生成器的输出应为元组`（x，y，sample_weight）`或`（x，y）`

1 个答案: