train_data = tf.keras.preprocessing.image_dataset_from_directory(DATA_DIR,
labels='inferred',
label_mode='categorical',
batch_size=BATCH_SIZE,
image_size=IMAGE_SIZE,
shuffle=True,
seed=SEED,
validation_split=0.2,
subset='training')
Found 600 files belonging to 3 classes.
Using 480 files for training.