tf_data_generator.py

import tensorflow as tf
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Flatten, Dense, Conv2D
from tensorflow.keras.preprocessing.image import ImageDataGenerator

train_datagen = ImageDataGenerator(
        rescale=1./255,
        validation_split=0.15
    )

train_generator = train_datagen.flow_from_directory(
            '../../Downloads/sss',
            target_size=(24, 24),
            batch_size=2,
            class_mode='binary')

for r in train_generator:
    print(r[0].shape, r[1].shape)
    print(r[0].dtype, r[1].dtype)
    break

data_generator = tf.data.Dataset.from_generator(lambda: train_generator, 
    output_types=(tf.float32, tf.float32),
    output_shapes=((2, 24, 24, 3), (2, )))

model = Sequential([
    Conv2D(10, (3, 3), activation='relu', input_shape=(24, 24, 3)),
    Flatten(),
    Dense(1, activation='sigmoid')
])

model.summary()

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

model.fit(data_generator, epochs=10, steps_per_epoch=1)

# model.fit_generator(
#     train_generator,
#     steps_per_epoch=10,
#     epochs=5
# )