Reputation: 1381
I'm looking for an efficient method to load PNG files as a TensorFlow 2.x date set. There are multiple answers here which all are using Tensorflow 1.x syntax. I also checked the official Tensorflow tutorial but they are using tf.keras.preprocessing.image_dataset_from_directory
API which is only available in tf-nightly.
Is there any API for Tensorflow 2.x capable of loading images from a folder, assigning the subfolder as the label, and return a TensorFlow dataset?
Upvotes: 0
Views: 126
Reputation: 101
I think you can use the flow_from_directory method. You might find this useful
Upvotes: 1