Hacker News new | ask | show | jobs
by emteycz 1550 days ago
No, the idea is that the model "builds" its own complex representations directly from images.