DALL-E
极客公园说
2021-01-06
8995 view · 89 likes
We’ve trained a neural network called DALL·E that creates images from text captions for a wide range of concepts expressible in natural language.



