I would strongly advice anyone interested by the labels frequencies of dataset to get a look at the json file provided.
Interesting thing is not that much the frequency of X or Y labels, but the frequency of one set of labels.
(this would be a great addition on the webpage actually).
Pictures can have multiple labels. And so, having the ratio of "Forum + Drugs + Finance" vs "Market-place + Weapons" dispense more information than just the global frequency of "Finance"-related pages :)
> We also manually removed pictures which were identified as containing harmful content, such as violent, offensive, obscene or equivalent undesirable pictures which may shock anyone.