I'm trying to train a CNN model with Keras.

For my data, i have images placed in my drive, when i use google colab, i access them with drive mount. When i use my local jupyter, i access them with my google drive application for desktop.

For getting the images, i'm using 'tf.keras.utils.image_dataset_from_directory' to run a CNN model in my Jupyter Anaconda environment. Here is the code:

base_dir = 'G:/Meu Drive/...'

batch_size_treino_val = 32

batch_size_teste = 1

def create_dataset(directory, batch_size):
    dataset = tf.keras.utils.image_dataset_from_directory(
        directory,
        labels='inferred',
        label_mode='binary',
        batch_size=batch_size,
        image_size=(150, 150),
        interpolation='bilinear',
        shuffle=True,
        class_names=['apneia', 'no_apneia']  # Especificando as classes desejadas
    )
    return dataset.prefetch(tf.data.AUTOTUNE)


train_ds = []
val_ds = []
test_ds = []
train_samples_count = 0
val_samples_count = 0
test_samples_count = 0

for patient in train_patients:
    path = os.path.join(base_dir, patient, 'segmentacao_8_segundos_sobreposicao_5_segundos', 'gaf')
    if os.path.exists(path):
        print(f"Processando: {path}")
        ds = create_dataset(path, batch_size_treino_val)
        num_images = sum(1 for _ in ds.unbatch())
        print(f"Imagens encontradas: {num_images}")

        train_size = int(0.8 * num_images)
        val_size = num_images - train_size

        train_samples_count += train_size
        val_samples_count += val_size

        train_ds.append(ds.take(train_size // batch_size_treino_val))
        val_ds.append(ds.skip(train_size // batch_size_treino_val))
    else:
        print(f"Diretório não encontrado: {path}")

But i'm getting the error:

---------------------------------------------------------------------------
NotFoundError                             Traceback (most recent call last)
Cell In[4], line 32
     30 if os.path.exists(path):
     31     print(f"Processando: {path}")
---> 32     ds = create_dataset(path, batch_size_treino_val)
     33     num_images = sum(1 for _ in ds.unbatch())
     34     print(f"Imagens encontradas: {num_images}")

Cell In[4], line 8, in create_dataset(directory, batch_size)
      6 def create_dataset(directory, batch_size):
----> 8     dataset = tf.keras.utils.image_dataset_from_directory(
      9         directory,
     10         labels='inferred',
     11         label_mode='binary',
     12         batch_size=batch_size,
     13         image_size=(150, 150),
     14         interpolation='bilinear',
     15         shuffle=True,
     16         class_names=['apneia', 'no_apneia']  # Especificando as classes desejadas
     17     )
     18     return dataset.prefetch(tf.data.AUTOTUNE)

File ~\anaconda3\lib\site-packages\keras\utils\image_dataset.py:210, in image_dataset_from_directory(directory, labels, label_mode, class_names, color_mode, batch_size, image_size, shuffle, seed, validation_split, subset, interpolation, follow_links, crop_to_aspect_ratio, **kwargs)
    208 if seed is None:
    209     seed = np.random.randint(1e6)
--> 210 image_paths, labels, class_names = dataset_utils.index_directory(
    211     directory,
    212     labels,
    213     formats=ALLOWLIST_FORMATS,
    214     class_names=class_names,
    215     shuffle=shuffle,
    216     seed=seed,
    217     follow_links=follow_links,
    218 )
    220 if label_mode == "binary" and len(class_names) != 2:
    221     raise ValueError(
    222         'When passing `label_mode="binary"`, there must be exactly 2 '
    223         f"class_names. Received: class_names={class_names}"
    224     )

File ~\anaconda3\lib\site-packages\keras\utils\dataset_utils.py:542, in index_directory(directory, labels, formats, class_names, shuffle, seed, follow_links)
    540 else:
    541     subdirs = []
--> 542     for subdir in sorted(tf.io.gfile.listdir(directory)):
    543         if tf.io.gfile.isdir(tf.io.gfile.join(directory, subdir)):
    544             if subdir.endswith("/"):

File ~\anaconda3\lib\site-packages\tensorflow\python\lib\io\file_io.py:768, in list_directory_v2(path)
    753 """Returns a list of entries contained within a directory.
    754 
    755 The list is in arbitrary order. It does not contain the special entries "."
   (...)
    765   errors.NotFoundError if directory doesn't exist
    766 """
    767 if not is_directory(path):
--> 768   raise errors.NotFoundError(
    769       node_def=None,
    770       op=None,
    771       message="Could not find directory {}".format(path))
    773 # Convert each element to string, since the return values of the
    774 # vector of string should be interpreted as strings, not bytes.
    775 return [
    776     compat.as_str_any(filename)
    777     for filename in _pywrap_file_io.GetChildren(compat.path_to_bytes(path))
    778 ]

NotFoundError: Could not find directory G:/Meu Drive/...

The fun fact is: When i do this exact same code in the Google Colab (replacing base_dir = '/content/drive/MyDrive... ) the code works and the tf.keras.utils.image_dataset_from_directory can find my images for training.

What am i doing wrong here?