Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

DINOv3

github.com

Post reply on HN
Print view View original

11–20 of 32 posts

Re: DINOv3

#11

Post by Qwuke »

llm_nerd wrote: post #10

You have to share your contact information, including DoB, and then be approved access, to obtain the models, and given that it's Meta I assume they're actually validating it against their All Humans database. They made their own DINOv3 license for this release (whereas DINOv2 used the Apache 2.0 license). Neat though. Will still check it out. As a first comment, I had to install the latest transformer==4.56.0dev (e.…

Yes, it's pretty disappointing for a seemingly big improvement over SOTA to be commercially licensed compared the previous version.. At least in the press release they're not portraying it as open source just because it's on GitHub/HuggingFace.

Re: DINOv3

#12

Post by pugworthy »

kaoD wrote: post #5
ranger_danger wrote: post #3

I have no idea what this even is.

> An extended family of versatile vision foundation models producing high-quality dense features and achieving outstanding performance on various vision tasks including outperforming the specialized state of the art across a broad range of settings, without fine-tuning

So, a group of AI models that can look at images and understand them, working well for many different tasks without needing extra training?

Re: DINOv3

#13

Post by tough »

Qwuke wrote: post #11
llm_nerd wrote: post #10

You have to share your contact information, including DoB, and then be approved access, to obtain the models, and given that it's Meta I assume they're actually validating it against their All Humans database. They made their own DINOv3 license for this release (whereas DINOv2 used the Apache 2.0 license). Neat though. Will still check it out. As a first comment, I had to install the latest transformer==4.56.0dev (e.…

Yes, it's pretty disappointing for a seemingly big improvement over SOTA to be commercially licensed compared the previous version.. At least in the press release they're not portraying it as open source just because it's on GitHub/HuggingFace.

The new Facebook AI Czar wang hinted on previous interviews that Meta might change their stand on licensing/open source.

Seems like the tides are shifting at meta

Re: DINOv3

#14

Post by rajman187 »

tough wrote: post #13
Qwuke wrote: post #11

Earlier quoted context omitted.

Yes, it's pretty disappointing for a seemingly big improvement over SOTA to be commercially licensed compared the previous version.. At least in the press release they're not portraying it as open source just because it's on GitHub/HuggingFace.

The new Facebook AI Czar wang hinted on previous interviews that Meta might change their stand on licensing/open source. Seems like the tides are shifting at meta

This has nothing to do with the newly appointed fellow nor Meta Superintelligence Labs, but rather work from FAIR that would have gone through a lengthy review process before seeing the light of day. Not fun to see the license change in any case

Re: DINOv3

#15

Post by stingraycharles »

pugworthy wrote: post #12
kaoD wrote: post #5

Earlier quoted context omitted.

> An extended family of versatile vision foundation models producing high-quality dense features and achieving outstanding performance on various vision tasks including outperforming the specialized state of the art across a broad range of settings, without fine-tuning

So, a group of AI models that can look at images and understand them, working well for many different tasks without needing extra training?

Exactly that. It’s easy to see why Meta, which Facebook and Instagram, is so heavily invested in these types of AI models.

Re: DINOv3

#16

Post by Imnimo »

I think SAM and DINO are the two off-the-shelf image models I've gotten the most mileage out of.

Re: DINOv3

#17

Post by kristopolous »

kaoD wrote: post #5
ranger_danger wrote: post #3

I have no idea what this even is.

> An extended family of versatile vision foundation models producing high-quality dense features and achieving outstanding performance on various vision tasks including outperforming the specialized state of the art across a broad range of settings, without fine-tuning

This is still pretty non-specific, however, luckily, they have a landing page:

https://ai.meta.com/dinov3/

Re: DINOv3

#18

Post by dennisloevlie »

rajman187 wrote: post #14
tough wrote: post #13

Earlier quoted context omitted.

The new Facebook AI Czar wang hinted on previous interviews that Meta might change their stand on licensing/open source. Seems like the tides are shifting at meta

This has nothing to do with the newly appointed fellow nor Meta Superintelligence Labs, but rather work from FAIR that would have gone through a lengthy review process before seeing the light of day. Not fun to see the license change in any case

I remember DINOv2 was originally a commercial licence. I (along with others) just asked if they could change it on a GitHub issue, and after some time, they did. Might be worth asking

Re: DINOv3

#19

Post by tough »

rajman187 wrote: post #14
tough wrote: post #13

Earlier quoted context omitted.

The new Facebook AI Czar wang hinted on previous interviews that Meta might change their stand on licensing/open source. Seems like the tides are shifting at meta

This has nothing to do with the newly appointed fellow nor Meta Superintelligence Labs, but rather work from FAIR that would have gone through a lengthy review process before seeing the light of day. Not fun to see the license change in any case

afaik now FAIR lives under the MSL umbrella https://x.com/AndrewCurran_/status/1940109829781270799
Post reply on HN