Keras creating three classes instead of two

问题

I am trying to train a model to identify images containing fire VS images that contain forests. I am training the model on a remote server using Linode. I am using Python 2.7 and Ubuntu 16.04.5.

When i run the following code locally or in Jupyter notebooks it will create 2 classes, but when i want to run it on the server it creates 3 classes.

The code that classifies the model:

def onehot(x): return np.array(OneHotEncoder().fit_transform(x.reshape(-1,1)).todense())


model = keras.applications.InceptionV3(weights='imagenet', include_top=False)


batch_size=16

train_datagen = ImageDataGenerator(
        rescale=1./255,
        shear_range=0.2,
        zoom_range=0.2,
        horizontal_flip=True)


train_generator = train_datagen.flow_from_directory(
        'datareal/train',  # this is the target directory
        batch_size=batch_size,
        target_size=(224, 224),
        class_mode='binary')

test_datagen = ImageDataGenerator(rescale=1./255)

validation_generator = test_datagen.flow_from_directory(
        'datareal/valid',
        batch_size=batch_size,
        target_size=(224, 224),
        class_mode='binary')


x = model.output
x = GlobalAveragePooling2D()(x)
# let's add a fully-connected layer
x = Dense(1024, activation='relu')(x)
# and a logistic layer -- let's say we have 2 classes
predictions = Dense(2, activation='softmax')(x)
newmodel = Model(inputs=model.input, outputs=predictions)

newmodel.compile(optimizer='rmsprop', loss='binary_crossentropy', metrics=['acc'])

Output:

Found 173 images belonging to 3 classes.
Found 40 images belonging to 3 classes.

The directory containing all the images is structured as follows:

datareal
        valid
                forest
                fire
        train
                forest
                fire

How do i get the model to label only 2 classes instead of 3?

回答1:

I figured it out. the problem was that i still had some hidden folders in my train and validation folder.

来源：https://stackoverflow.com/questions/52609798/keras-creating-three-classes-instead-of-two

标签

python

python-2.7

tensorflow

keras

linode