TF, access the filenames of the elements in a dataset object: dataset.group_by_window group by filename?

Solution 1:

Use the property file_paths to the access the file names of your dataset:

import tensorflow as tf
import pathlib

dataset_url = "https://storage.googleapis.com/download.tensorflow.org/example_images/flower_photos.tgz"
data_dir = tf.keras.utils.get_file('flower_photos', origin=dataset_url, untar=True)
data_dir = pathlib.Path(data_dir)

batch_size = 32

train_ds = tf.keras.utils.image_dataset_from_directory(
  data_dir,
  validation_split=0.2,
  subset="training",
  seed=123,
  image_size=(180, 180),
  batch_size=batch_size)

def preprocess_data(images, labels):
  file_paths = train_ds.file_paths
  # Do something with the file_paths
  # ...

  print(file_paths[:5])
  return images, labels
  
train_ds = train_ds.map(preprocess_data)
# or train_ds.group_by_window(*)

Found 3670 files belonging to 5 classes.
Using 2936 files for training.
['/root/.keras/datasets/flower_photos/daisy/7568630428_8cf0fc16ff_n.jpg',
 '/root/.keras/datasets/flower_photos/dandelion/7367491658_9eb4dc2384_m.jpg',
 '/root/.keras/datasets/flower_photos/tulips/18378582936_ee7085c850.jpg', 
 '/root/.keras/datasets/flower_photos/tulips/2426847695_4b8409402e_n.jpg', 
 '/root/.keras/datasets/flower_photos/roses/2949945463_366bc63079_n.jpg']