Thanks, I think I understand roughly. Could it be used for for recognizing people? As in identifying what person is in what image?

Caveat: I am not an expert, so this is a semi-educated guess.

I imagine it would depend on whether DINOv3 captures the information of whether a given person is in the image, which I think is really a question about training data. So naively, I would guess the answer is yes for celebrities and no for non-celebrities. Partially for data/technical reasons, but also maybe due to the murkier legal expectation of privacy for famous people.

Foundation models like DINO learn representations of their inputs. That is, they generate very high-dimensional numerical descriptions of what you put into them. The models aren't trained on labelled data, but they're trained on some pretext task like "given this image with a cutout, fill in the cutout" (see Masked Auto-Encoders). So the basic output from a model is a vector - often called an embedding. Literally a 1D list of numbers, O(1k)-dimensional. Your goal is to get an embedding that assigns (well) linearly separable vectors for all the things you want to classify.

Vision transformers also output patch tokens, which can be assembled into a low-resolution feature map (w/32, h/32 is common). So what you do with that data depends on the task. Classification can be as simple as linearly classifying the (whole image) embedding. A semantic segmentation task can do the same, but for every pixel. This is why the DINO authors show a PCA representation of a bunch of images, which show that semantically similar objects are grouped together by colour. Object detectors are more complicated, but the key thing is that once you have these pixel-level features, you can use them as input into existing architectures.

Now to your question: face recognition is a specific application of object re-identification (keyword: Re-ID). The way most of these models work is from the whole-image embedding. Normally you'd run a detector to extract the face region, then compute the embedding, put it in a vector database and then query for nearest neighbours using something like the cosine distance. I've only worked in this space for animals, but humans are far more studied. Whether DINOv3 is good enough out-of-the-box I don't know, but certainly there's a lot of literature looking at these sorts of models for Re-ID.

The challenge with Re-ID is that the model has to be able to produce features which discriminate individuals rather than similar looking individuals. For example with the vanilla model, you probably have a very good tool for visual search. But that's not the same task, because if you give it a picture of someone in a field, you'll get back pictures of other people in fields. That usually requires re-training on labelled imagery where you have a few examples of each person. The short answer is that there are already very good models for doing this, and they don't necessarily even need ML to do a decent job (though it might be used for keypoint detection for facial landmarks).

I'm not confident in what I'm saying here, so please correct me if I'm wrong as I'd like to learn:

Human hearing isn't linear in terms of loudness. So a 3db increase in loudness sounds like "an increase", but the pressure is actually double. Hence, it makes sense to use db to describe loudness even in the context of perceived loudness to human-hearing.

This is similar to brightness. In photography, "stops" are used to measure brightness. One stop brighter is technically twice the light, but to the human eye, it just looks "somewhat brighter", as human brightness appreciation is logarithmic, just like "stops" and "db".

Thank you, I'm already paying for a lower-tier Kagi plan. Forgot about their AI offering.

Excellent, thank you

I _need_ to tinytapout! Don't know what, don't know how, but the sense of need is tickling! I feel like twelve again in front of a 286 without a case.

Would it be possible for the allocator/GC to know what allocations are made within a request and make a generation for specifically for it? Allocations too big to fit would be made like usual

That's already what we effectively have.

Since objects cannot be promoted to the old generation inside the request cycle, objects in the new gen are request allocated objects.

So if we were to eagerly trigger a minor GC after a request, we'd have very little objects to scan, and only need to sweep garbage, which is only a small fraction of time spent in GC.

This is why I come to HN. Thank you to the author.

Very nice, thanks for sharing. I have two questions relevant to me:

- Will the purchase include upgrades? - I have most of my media on a NAS. Will this app be useful in my case? Would I be indexing a network drive?

Thanks!

1) Yes the purchase includes upgrades. After you install those happen automatically.

2) This doesn't support NAS yet, but it's on the roadmap. If enough people are interested in that I'd bump it up in priority.

i'm definitely interested in that

The upgrade is particularly relevant for the Apple ecosystem where updates and API changes regularly weed out unmaintained apps.

Upgrades are included in the purchase! Once you install Desktop Docs you'll get the new versions automatically.

Here's an interesting video about a type of Maize that can also create its own "bound" Nitrogen https://www.youtube.com/watch?v=CFyd-kC6IUw

Ah, that's a relief. It sounded incredibly scary if it was some new type of species just now. I would imagine that it would overwhelm our ecosystems wreaking havoc until a new balance is eventually found with new winners and new types of species dominating our environment.