EditAnything: Segment Anything + ControlNet + BLIP2 + Stable Diffusion

Re:EditAnything: Segment Anything + ControlNet + BLIP2 + Sta... by 55555 on Monday April 10, 2023 @11:22AM (#35513410)

Where did META get the Segment Anything training dataset from? Or, if they made it, how did they generate such a massive dataset of accurate masks?

[ Parent | on HN ]

Re:EditAnything: Segment Anything + ControlNet + BLIP2 + Sta... by xiphias2 on Monday April 10, 2023 @11:36AM (#35513591)

Actually it is quite small dataset by current standards (millions of images), so I guess it was using humans in the loop.

[ Parent | on HN ]

Re:EditAnything: Segment Anything + ControlNet + BLIP2 + Sta... by fkarg on Monday April 10, 2023 @12:31PM (#35514152)

the relevant part is the masks, and so far it's the biggest dataset in existence, with about 400x more masks than the next smaller one.

: excluding non-public ones

[ Parent | on HN ]

Re:EditAnything: Segment Anything + ControlNet + BLIP2 + Sta... by fkarg on Monday April 10, 2023 @12:29PM (#35514136)

they described it in their paper: they created it themselves. they had/have a loop of having humans annotate images their system was uncertain about, but the released one was fully automated.

[ Parent | on HN ]