Some tasks are inferred based on the benchmarks list.
The benchmarks section lists all benchmarks using a given dataset or any of
its variants. We use variants to distinguish between results evaluated on
slightly different versions of the same dataset. For example, ImageNet 32⨉32
and ImageNet 64⨉64 are variants of the ImageNet dataset.
IG-3.5B-17k is an internal Facebook AI Research dataset for training image classification models. It consists of hashtags for up to 3.5 billion public Instagram images.