2 ms·
I think the training data is just image/caption pairs. I don't think there's any notion of localizing or detecting specific objects in the training images.
by Imnimo 4y ago
I think the training data is just image/caption pairs. I don't think there's any notion of localizing or detecting specific objects in the training images.