'predict_generator' return values over then 1 and less then 0

Question

I used Autokeras to train my model, after then fit_final I save it as pure keras h5 file.

My model:

from autokeras import ImageClassifier
from autokeras.image.image_supervised import load_image_dataset

if __name__ == '__main__':
    x_test, y_test = load_image_dataset(csv_file_path="test/label.csv", images_path="test")
    print(x_test.shape)
    print(y_test.shape)

    x_train, y_train = load_image_dataset(csv_file_path="train/label.csv", images_path="train")
    print(x_train.shape)
    print(y_train.shape)

    clf = ImageClassifier(path="~/automodels/", verbose=True)
    clf.fit(x_train, y_train, time_limit= 1 * 10 * 60)
    clf.final_fit(x_train, y_train, x_test, y_test, retrain=True)
    y = clf.evaluate(x_test, y_test)
    print(y)

    clf.export_autokeras_model('my_autokeras_model.h5ak')
    clf.export_keras_model('my_model.h5')

I also have a predict.py code but it gives me wrong values

from keras.models import load_model
from keras.preprocessing import image
import numpy as np
import glob
from keras.preprocessing.image import ImageDataGenerator
from sklearn.metrics import confusion_matrix

# dimensions of our images
img_width, img_height = 128, 128

# load the model we saved
model = load_model('model.h5')
#model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])

datagen = ImageDataGenerator(rescale=1./255)

generator = datagen.flow_from_directory(
        'data/test',
        target_size=(img_width, img_height),
        batch_size=1,
        class_mode=None,  # only data, no labels
        shuffle=False)  # keep data in same order as labels

#filenames = datagen.filenames
#nb_samples = len(filenames)

probabilities = model.predict_generator(generator, 4)

actual result:

[[-2.0996048  1.862035 ]
 [-1.4634153  1.2710633]
 [-1.4367918  1.4041075]
 [-1.3242773  1.2946494]]

expected result should look like these:

[[0  0.51234 ]
 [1  0.67847]
 [1  0.92324]
 [1  0.32333]]

for example.

What am I doing wrong?

Answer 1

After Mickey adviced me about the activation function, I found this thread on github here

this thread helped me to figure out those lines of code:

keras_model = load_model('model.h5')
x = keras_model.output
x = Activation('softmax', name='activation_add')(x)
new_model = Model(keras_model.input, x)