Dog noses are individual like fingerprints – mind the implications

[Ruben]

Dog noses can be used to identify dogs similarly to how human fingerprints are used to identify humans. And it's put into practice now:

NOSEiD app helps reunite lost dogs with their owners using photos of their noses

The dog food company, Iams has launched a new app, NOSEiD, that photographs and scans dog noses to identify lost pups and help reconnect them with their owners.

www.dpreview.com

That's cool when it is what helps you find a lost dog, but think about the implications in the near future. Photos/videos of your dogs' noses of sufficient quality can be automatically matched with photos of your dogs' noses elsewhere (social media) and of course with a database in case your dogs' noses have been registered either by yourself or by previous owners.

So be careful. It's another reason not to share content, if it is not allowed where you live.

 

[Deleted member]

I vote this be added to some pinned thread

 

[Jewelry Fish]

Dog noses can be used to identify dogs similarly to how human fingerprints are used to identify humans. And it's put into practice now:

NOSEiD app helps reunite lost dogs with their owners using photos of their noses

The dog food company, Iams has launched a new app, NOSEiD, that photographs and scans dog noses to identify lost pups and help reconnect them with their owners.

www.dpreview.com

That's cool when it is what helps you find a lost dog, but think about the implications in the near future. Photos/videos of your dogs' noses of sufficient quality can be automatically matched with photos of your dogs' noses elsewhere (social media) and of course with a database in case your dogs' noses have been registered either by yourself or by previous owners.

So be careful. It's another reason not to share content, if it is not allowed where you live.

Thanks for this, Ruben. That’s some important info for future posters

 

[FoxandtheHounds]

Yikes, thanks for posting this. Good to know!

 

[Labraguy5]

I can just see it in a few years time...how to spot a zoo on Facebook

But great post! Same goes for unique markings or features.

 

[Pillar]

Yow, they can track your dog's sniffer now?

 

[Deleted member]

Yow, they can track your dog's sniffer now?

We don't know if zealots will convince LE to use manpower and resources to "indefinately I don't know of a possible maybe" expose someone doing their mutt. But in a day and age where a tonne of cops threatened deadly force against a dude peacefully meditating on a beach that was closed, I wouldn't put it past them.

 

[zooms]

Dog noses can be used to identify dogs similarly to how human fingerprints are used to identify humans. And it's put into practice now:

NOSEiD app helps reunite lost dogs with their owners using photos of their noses

The dog food company, Iams has launched a new app, NOSEiD, that photographs and scans dog noses to identify lost pups and help reconnect them with their owners.

www.dpreview.com

That's cool when it is what helps you find a lost dog, but think about the implications in the near future. Photos/videos of your dogs' noses of sufficient quality can be automatically matched with photos of your dogs' noses elsewhere (social media) and of course with a database in case your dogs' noses have been registered either by yourself or by previous owners.

So be careful. It's another reason not to share content, if it is not allowed where you live.

The quality needs to be good and there must be focus on the nose for it to work but being cautious is good, it's the same for your hands

 

[Pillar]

We don't know if zealots will convince LE to use manpower and resources to "indefinately I don't know of a possible maybe" expose someone doing their mutt. But in a day and age where a tonne of cops threatened deadly force against a dude peacefully meditating on a beach that was closed, I wouldn't put it past them.

Whoa, was this in China or America?

 

[Austral]

Wow, I thought that would be something really good, until I read your last sentence

Still, implementing that in the app seems pretty useless, it's already enough with the adds from both sides, the location and a picture.

But maybe this could be used to prosecute zoos (and not essentially the police, but haters), using a software similar to make a database from pics from social media and comparing them with zoo material.

 

[dartel]

Really doesn't seem like there's any danger there. Looks like you need a close up of just the nose to "register" your dog into the database, then if a lost dog is found the finder would take another closeup that the app would print against the database. A wide shot of even the full head wouldn't have enough resolution to ID against. So there wouldn't be people finding a zoo image online and comparing it against facebook pet photos to identify the zoo.

 

[Deleted member]

Whoa, was this in China or America?

Florida, USofA

 

[Ruben]

The quality needs to be good and there must be focus on the nose for it to work but being cautious is good, it's the same for your hands

Indeed.
Be extra careful with videos though. If you pause a video and look at the still image, what you see will be less than the detail that can actually be extracted from multiple frames. It's possible to extract sub-pixel data from videos (which are not smoothed by compression too much).

 

[Labraguy5]

Really doesn't seem like there's any danger there. Looks like you need a close up of just the nose to "register" your dog into the database, then if a lost dog is found the finder would take another closeup that the app would print against the database. A wide shot of even the full head wouldn't have enough resolution to ID against. So there wouldn't be people finding a zoo image online and comparing it against facebook pet photos to identify the zoo.

Luckily with social media, storage equals less profit so they compress the shit out of everything, so it would have to be a clear, close up face pic of your dog. The latest stat I could find is that Facebook gets 350 million photo uploads a day, so scanning that many photos for one dog nose wouldn't be easy and need a whole lot of resource, although not impossible.

 

[Ruben]

The latest stat I could find is that Facebook gets 350 million photo uploads a day, so scanning that many photos for one dog nose wouldn't be easy and need a whole lot of resource, although not impossible.

That's not how it would be done. You'd scan all the photos only once and save the characteristic pattern of the dog noses in the form of a bunch of numbers inside a database. Then, each time you want to identify a particular dog, you only need to analyze its – that is a single – photo, calculate the numbers that correspond to this dog's nose and search for these in your pre-made database.

The search does not even require you to compare the numbers corresponding to the wanted dog with hundreds of millions other entries, although even that could be done within fractions of a second. The database doesn't need to be a linear list. It can branch out as a tree-like structure at each characteristic that is saved about noses. For example, if a dog has a brown nose, the search won't have to go through any of the hundreds of millions black noses at all. It only needs to compare with other brown noses. And this is in principle true for any feature of the nose pattern. A few dozen comparisons may be enough to identify one nose among hundreds of millions like this.

 

[Labraguy5]

That's not how it would be done. You'd scan all the photos only once and save the characteristic pattern of the dog noses in the form of a bunch of numbers inside a database. Then, each time you want to identify a particular dog, you only need to analyze its – that is a single – photo, calculate the numbers that correspond to this dog's nose and search for these in your pre-made database.

The search does not even require you to compare the numbers corresponding to the wanted dog with hundreds of millions other entries, although even that could be done within fractions of a second. The database doesn't need to be a linear list. It can branch out as a tree-like structure at each characteristic that is saved about noses. For example, if a dog has a brown nose, the search won't have to go through any of the hundreds of millions black noses at all. It only needs to compare with other brown noses. And this is in principle true for any feature of the nose pattern. A few dozen comparisons may be enough to identify one nose among hundreds of millions like this.

I totally agree, but I meant currently with no system in place they would just have to mass scan an insane amount of photos. And these companies are all about profit, if there's little benefit they won't invest the money for development and then running cost like server capacity. Aside from the risk to us, it's a shame as dog theft is on the rise, and a system like this would be really good to track down stolen loved ones.

 

[Ruben]

I totally agree, but I meant currently with no system in place they would just have to mass scan an insane amount of photos. And these companies are all about profit, if there's little benefit they won't invest the money for development and then running cost like server capacity. Aside from the risk to us, it's a shame as dog theft is on the rise, and a system like this would be really good to track down stolen loved ones.

I could imagine that a company such as Facebook will buy or replicate the tech and implement it in their website automatically then. They can identify people in uploaded photos automatically already, don't they?

 

[Labraguy5]

I could imagine that a company such as Facebook will buy or replicate the tech and implement it in their website automatically then. They can identify people in uploaded photos automatically already, don't they?

It's possible, but it still costs money to buy and run a service like that and they will only do that either if users demand it or it makes them more money. For example on the scale, they still need to open every newly uploaded photo and check if there's a dog in it before profiling. If that takes a server 0.5 seconds per photo, then times that by around 4000 photo uploads a second, they'd need 2000 servers which need power, A/C and maintenance. Then there's the billions of existing photos which they already have. I think currently they've got bigger problems than outing a few zoos, like child grooming and the likes on their platform, but maybe in the far future it's a possibility.

The facial recognition for humans was a fail, it was really hit and miss or tagging the wrong people and most users opted out as no one wanted it (plus it was creepy).

 

[Ruben]

It's possible, but it still costs money to buy and run a service like that and they will only do that either if users demand it or it makes them more money. For example on the scale, they still need to open every newly uploaded photo and check if there's a dog in it before profiling. If that takes a server 0.5 seconds per photo, then times that by around 4000 photo uploads a second, they'd need 2000 servers which need power, A/C and maintenance. Then there's the billions of existing photos which they already have. I think currently they've got bigger problems than outing a few zoos, like child grooming and the likes on their platform, but maybe in the far future it's a possibility.

The facial recognition for humans was a fail, it was really hit and miss or tagging the wrong people and most users opted out as no one wanted it (plus it was creepy).

A company like Facebook is interested to identify what exactly is in a photo for different reasons than outing zoos. It's the same with search engines like Google. Data is what they capitalize on. The possibility to out zoos would be merely a side effect that other people could use, either law enforcement or even "hobby detectives" then.

 

[dartel]

That's not how it would be done. You'd scan all the photos only once and save the characteristic pattern of the dog noses in the form of a bunch of numbers inside a database. Then, each time you want to identify a particular dog, you only need to analyze its – that is a single – photo, calculate the numbers that correspond to this dog's nose and search for these in your pre-made database.

The search does not even require you to compare the numbers corresponding to the wanted dog with hundreds of millions other entries, although even that could be done within fractions of a second. The database doesn't need to be a linear list. It can branch out as a tree-like structure at each characteristic that is saved about noses. For example, if a dog has a brown nose, the search won't have to go through any of the hundreds of millions black noses at all. It only needs to compare with other brown noses. And this is in principle true for any feature of the nose pattern. A few dozen comparisons may be enough to identify one nose among hundreds of millions like this.

Exactly. It wouldn't be facebook/whatever doing it. It would be armchair vigilantes plugging zoo photos into the app then trying to use it to match to non-zoo photos to rat them out. But yeah, it wouldn't have the resolution to work that way. From the looks of it you need a closeup nose only shot, plugging in a long shot of the full head/body/scene wouldn't work. That would be like taking a picture of someone's hand from 10 feet away and trying to use it for fingerprinting, not gonna work.

 

[zooms]

Is artificial intelligence unheard of in the zoophile community, that would be more likely to be used then a database of pictures

 

[Pillar]

Florida, USofA

Ah, there. I don't know if you're in the US but we've been having this problem for decades and we haven't been holding these freaks with badges accountable frequently enough...

Oh and BTW the governor there is a fascist pig.