Keras setting up dataset - tensorflow

Question

I am trying to setup my train, validation, and test data and am running into an error:
Traceback (most recent call last):
line 28, in <module>
train_data_gen = train_image_generator.flow_from_directory(directory=data_directory,
, line 958, in flow_from_directory
return DirectoryIterator(
line 376, in __init__
super(DirectoryIterator, self).__init__(
, line 113, in __init__
if not classes:
ValueError: The truth value of an array with more than one element is ambiguous. Use a.any()
or a.all()
The code I have so far is:
import pathlib
import numpy as np
from tensorflow.keras.preprocessing.image import ImageDataGenerator
# Process images
# initializing some training settings
im_height = 224
im_width = 224
batch_size = 32
epoch = 50
# assigning directory for images
data_directory = pathlib.WindowsPath(".../Desktop/Data/fruit-dataset/train")
test_data = pathlib.WindowsPath(".../Desktop/Data/fruit-dataset/test")
# assigning classes based off folder structure
CLASSES = np.array([item.name for item in data_directory.glob('*') if item.name != "LICENSE.txt"])
# image normalization with selected augmentation settings
train_image_generator = ImageDataGenerator(rescale=1 / 255.0,
horizontal_flip=True,
validation_split=0.20
)
train_data_gen = train_image_generator.flow_from_directory(directory=data_directory,
batch_size=batch_size,
shuffle=True,
target_size=(im_height, im_width),
classes=CLASSES,
class_mode='categorical',
color_mode='rgb'
)
test_data_gen = ImageDataGenerator(rescale=1 / 255.0)
test_generator = test_data_gen.flow_from_directory(directory=test_data,
target_size=(im_height, im_width),
color_mode="rgb",
batch_size=batch_size,
class_mode="categorical",
shuffle=False,
)
I am also new to using TF and am unsure how I would display a batch to my screen to further ensure I'm loading the data.

Answer 1

make sure your train and test directories only contain the six sub directories. No files. In each of the sub directories make sure all the files are image files. No non-image files. Forget about using the code
CLASSES = np.array([item.name for item in data_directory.glob('*') if item.name != "LICENSE.txt"])
leave classes=None in the generators.
flow_from_directory will automatically assign class names based on the names of the sub directories. Make sure the sub directories have EXACTLY the same names in both the training directory and the test directory.