Question

我按照this link的指南建立了一个模型，并在微调部分之前停止使用以下代码在其他一些图像上测试模型：

img_width, img_height = 150, 150
batch_size = 1

test_model = load_model('dog_cat_model.h5')

validation_data_dir = "test1"

test_datagen = ImageDataGenerator(rescale=1. / 255)

validation_generator = test_datagen.flow_from_directory(
    validation_data_dir,
    target_size=(img_width, img_height),
    batch_size=batch_size,
    shuffle=False,
    class_mode='binary')

predictions = test_model.predict_generator(validation_generator, len(validation_generator.filenames));

for i in range(len(validation_generator.filenames)):
    print(validation_generator.filenames[i], ": ", predictions[i])

但是我收到以下错误：

ValueError: Error when checking : expected flatten_1_input to have shape (None, 4, 4, 512) but got array with shape (1, 150, 150, 3)

打印test_model.summary提供以下输出：

_________________________________________________________________
Layer (type)                 Output Shape              Param #
=================================================================
flatten_1 (Flatten)          (None, 8192)              0
_________________________________________________________________
dense_1 (Dense)              (None, 256)               2097408
_________________________________________________________________
dropout_1 (Dropout)          (None, 256)               0
_________________________________________________________________
dense_2 (Dense)              (None, 1)                 257
=================================================================
Total params: 2,097,665
Trainable params: 2,097,665
Non-trainable params: 0
_________________________________________________________________
None

我无法弄清楚这意味着什么。

以下是我用于创建模型的代码：

img_width, img_height = 150, 150
top_model_weights_path = 'bottleneck_fc_model.h5'
train_data_dir = 'data/train'
validation_data_dir = 'data/validation'
batch_size = 16
train_samples = 2000
validation_samples = 800
epochs = 50


def save_bottlebeck_features():
    datagen = ImageDataGenerator(rescale=1. / 255)

    # build the VGG16 network
    model = applications.VGG16(include_top=False, weights='imagenet')

    train_generator = datagen.flow_from_directory(
        train_data_dir,
        target_size=(img_width, img_height),
        batch_size=batch_size,
        class_mode=None,
        shuffle=False)

    validation_generator = datagen.flow_from_directory(
        validation_data_dir,
        target_size=(img_width, img_height),
        batch_size=batch_size,
        class_mode=None,
        shuffle=False)

    predict_size_train = int(math.ceil(train_samples / batch_size))

    bottleneck_features_train = model.predict_generator(train_generator, predict_size_train)
    np.save('bottleneck_features_train.npy', bottleneck_features_train)

    predict_size_validation = int(math.ceil(validation_samples / batch_size))

    bottleneck_features_validation = model.predict_generator(validation_generator, predict_size_validation)
    np.save('bottleneck_features_validation.npy', bottleneck_features_validation)


def train_top_model():
    train_data = np.load('bottleneck_features_train.npy')
    train_labels = np.array([0] * (train_samples // 2) + [1] * (train_samples // 2))

    validation_data = np.load('bottleneck_features_validation.npy')
    validation_labels = np.array([0] * (validation_samples // 2) + [1] * (validation_samples // 2))

    model = Sequential()
    model.add(Flatten(input_shape=train_data.shape[1:]))
    model.add(Dense(256, activation='relu'))
    model.add(Dropout(0.5))
    model.add(Dense(1, activation='sigmoid'))

    model.compile(optimizer='rmsprop',
                  loss='binary_crossentropy',
                  metrics=['accuracy'])

    model.fit(train_data, train_labels,
              epochs=epochs,
              batch_size=batch_size,
              validation_data=(validation_data, validation_labels))

    model.save_weights(top_model_weights_path)
    model.save('dog_cat_model.h5')


save_bottlebeck_features()
train_top_model()

我希望有人可以帮助我:)

Answer 1

尝试通过GlobalAveragePooling2D或GlobalMaxPooling2D替换Flatten图层

Answer 2

您在此行中混合了bootleneck和图像模型逻辑：

test_model.predict_generator(...)

将图像输送到瓶颈模型的位置。这导致了形状错误。

Answer 3

添加

ss.CharAt(i - 1)

data_format='channels_last'

中的

如果后端使用Tensorflow，则

将起作用，

model.add()

如果后端正在使用theano。

ValueError：检查时出错：预期flatten_1_input具有形状（无，4,4,512）但是具有形状的数组（1,150,150,3）

3 个答案: