I want to use ImageNet (ILSVRC) to make hierarcical classification by leveraging the hierarcal structure of ImageNet.
In Cifar100 for example, each image is associated with a corase and a fine label, which represent levels of a hierarchy.
How can I have such a hierarchy in ImageNet using tensorflow/Keras?
I have downloaded ImageNet data, and as I can see, there is 1 folder for each of the 1000 classes.
How to know to which of the supposed 2,155 nodes each image sample is associated to? And how to load that information in tensorflow? Thanks