Fake views for the win: Text-to-image models learn more efficiently with made-up data – The Register
Fake views for the win: Text-to-image models learn more efficiently with made-up data – The Register