I'm currently using the coco/2017 dataset for some use in tf.keras transferred learning
I am aware that you can download the full dataset using the code bellow:
(raw_train, raw_validation, raw_test), metadata = tfds.load(
'coco/2017',
split=list(splits),
with_info=True,
as_supervised=True)
How would I do this if I only wanted to extract the 'persons' images and labels. Instead of the full dataset?