Z9 'Deep-learning AI'

[TonyF]

There doesn't seem to be much said or written about this. Apart from advanced subject detection, does anyone have more insight as to what this is all about.
Marketing say - 'Deep-learning AI, meanwhile, powers the most advanced, simultaneous detection of up to nine different subject types'
What's your interpretation?

 

[NorthernFocus]

There doesn't seem to be much said or written about this. Apart from advanced subject detection, does anyone have more insight as to what this is all about.
Marketing say - 'Deep-learning AI, meanwhile, powers the most advanced, simultaneous detection of up to nine different subject types'
What's your interpretation?

Resistance is futile....

 

[fcotterill]

some context...

Clearing the Confusion: AI vs Machine Learning vs Deep Learning Differences

Raise your hand if you’ve been caught in the confusion of differentiating artificial intelligence (AI) vs machine learning (ML) vs deep learning (DL)… Although the three terminologies are usually…

towardsdatascience.com

Machine Learning for Photographers: Image Tracking - ERIC KIM

Basic idea: as a photographer, if you're shooting street photography (or photographing anything that is moving), you are tracking the movement of the subject. The movement of the subject moving towards you, away from you, or in different directions. But how can we apply this machine learning...

erickimphotography.com

Nikon invested in startups and more in AI as part of its strategic embracing of robotics

Nikon invests $7.5 million in computer vision and deep learning startup Wrnch – Robotics & Automation News

roboticsandautomationnews.com

 

[Maljo]

Here is what we know so far…. It’s. UnStoppable.

 

[simonsi]

There doesn't seem to be much said or written about this. Apart from advanced subject detection, does anyone have more insight as to what this is all about.
Marketing say - 'Deep-learning AI, meanwhile, powers the most advanced, simultaneous detection of up to nine different subject types'
What's your interpretation?

Pretty sure that any learning of any kind is done off-camera and only the results are built in to the camera itself, your camera doesn't have access to much data except what you shoot and it isn't being told what the subject actually is so could happily call your bird in flight a race car. This would be the same as the matrix meter, the development of subject models happens in development and the camera meter can then refer to the library of subjects when metering.

 

[Nimi]

Pretty sure that any learning of any kind is done off-camera and only the results are built in to the camera itself, your camera doesn't have access to much data except what you shoot and it isn't being told what the subject actually is so could happily call your bird in flight a race car. This would be the same as the matrix meter, the development of subject models happens in development and the camera meter can then refer to the library of subjects when metering.

You think they continue to compile images and future firmware would incorporate it, or is it etched into the processor?

 

[simonsi]

You think they continue to compile images and future firmware would incorporate it, or is it etched into the processor?

the libraries of data can be updated with firmware I’d imagine but they could equally be hard coded in rom for speed.

either way the “powered by deep learning” is marketing spin, the individual camera isnt doing any learning

 

[NorthernFocus]

...the individual camera isnt doing any learning

I'm not so sure about that. I wake up some nights and hear strange noises coming from the camera locker. I'm careful not to leave that thing mounted on a tripod overnight. Who knows what it would do if it had legs

 

[Nimi]

the libraries of data can be updated with firmware I’d imagine but they could equally be hard coded in rom for speed.

either way the “powered by deep learning” is marketing spin, the individual camera isnt doing any learning

Is that true for phones too, or do they have enough processing power and/or access to it via cell transmission (or wifi) to learn?

 

[John Navitsky]

the individual camera isnt doing any learning

the a1 has a face registration thingy. prolly mostly a gimmick, but maybe some limited stuff might be possible

 

[TonyF]

some context...

Clearing the Confusion: AI vs Machine Learning vs Deep Learning Differences

Raise your hand if you’ve been caught in the confusion of differentiating artificial intelligence (AI) vs machine learning (ML) vs deep learning (DL)… Although the three terminologies are usually…

towardsdatascience.com

Machine Learning for Photographers: Image Tracking - ERIC KIM

Basic idea: as a photographer, if you're shooting street photography (or photographing anything that is moving), you are tracking the movement of the subject. The movement of the subject moving towards you, away from you, or in different directions. But how can we apply this machine learning...

erickimphotography.com

Nikon invested in startups and more in AI as part of its strategic embracing of robotics

Nikon invests $7.5 million in computer vision and deep learning startup Wrnch – Robotics & Automation News

roboticsandautomationnews.com

Thank You. I can now see that the use of the term 'deep-learning AI' is somewhat legitimate because it is a subset of AI with the ability to go beyond machine learning and automatically recognise subjects.

I can also stop overthinking this and not worry about any need to train my camera

 

[Woodpecker]

FWIW

With all that we should never forget that marketing people go for this kind of buzz terms like insects go for UV lamps, while 90% of them don't even know what they are talking about. My professional world is process automation and MOM (Manufacturing Operations Management) and is was and still is amazing to see what was called AI, machine learning, deep learning ... you name it ... turned out to be a truly complex and truly well-designed - but nevertheless - classic algorithms camouflaged as "intelligence", while the people designing these are still by far more intelligent compared to what the marketing people call AI in whatever colour or shape.

The moment we create the first machine that is really intelligent - or at least more that good human brain, this will pretty shortly turn out to be the last invention that was first made by humans.

Perfect technical support is - no doubt - really great to have, but the moment the tool is becoming more intelligent than the worker, you are not very far away from the worker being redundant. As all of us I am sad if I missed a great shot, especially in a once-in-a-lifetime situation, but the question is whether it is the image that counts or the moment in which I have encountered the scene with all my senses. Don't we all - well, at leat most of us - still tend to get more satisfaction from an achievement if it was a difficult one ???