How good a camera is the Z9 really?

If you would like to post, you'll need to register. Note that if you have a BCG store account, you'll need a new, separate account here (we keep the two sites separate for security purposes).

How good is the Z9? Well, it is excellent.

I don't see the limitations that others describe on this forum. But, I don't shoot BIF. I mainly shoot humans.

With all the problems described here I recently have been trying to see if I can replicate these focus searching problems for the event and performance use cases and I cannot.

It can shoot precisely focused eyes of moving musicians through a maze of cables, microphone stands and flying drumsticks. It can pick individual eyes out of a jostling gaggle of people. All this often at above ISO 3200 at f2.8 or even f1.2.

Whereas with previous cameras I would need to push a single-point focus spot around the frame to achieve a desired composition (with many lost opportunities) , the Z9 finds and tracks the eye automatically. Or, locks on with 3D tracking if there are choices.

For humans I find the Z9 hard to stump.

Makes me wonder if the training set for eye detect for humans was just better than the one for animals.
 
Makes me wonder if the training set for eye detect for humans was just better than the one for animals.
as someone who primarily shoots dogs, i’ve long ago concluded animals are a super difficult target compared to humans. it took me many years to build enough knowledge and skill to do it semi-reliability with the d500.

dogs for example are super fast, unpredictable, they are relatively small, their head even smaller, they come in all sorts of different colors and color patterns and textures and their fur is much less refractive than skin and some super unrefractive.

it’s just a harder problem, imo

i will point out that out of the box the z9 did a great job with dogs playing disc. my worries about committing to the z9 instantly went away
 
How good is the Z9? Well, it is excellent.

I don't see the limitations that others describe on this forum. But, I don't shoot BIF. I mainly shoot humans.

With all the problems described here I recently have been trying to see if I can replicate these focus searching problems for the event and performance use cases and I cannot.

It can shoot precisely focused eyes of moving musicians through a maze of cables, microphone stands and flying drumsticks. It can pick individual eyes out of a jostling gaggle of people. All this often at above ISO 3200 at f2.8 or even f1.2.

Whereas with previous cameras I would need to push a single-point focus spot around the frame to achieve a desired composition (with many lost opportunities) , the Z9 finds and tracks the eye automatically. Or, locks on with 3D tracking if there are choices.

For humans I find the Z9 hard to stump.

Makes me wonder if the training set for eye detect for humans was just better than the one for animals.

I don't know about the Z9 specifically but I agree with your observation for all the brands that I have shot for street vs wildlife photography.
I suspect there are quite a few reasons behind this - first human face detection and eye detection has been in development for a much longer time and I believe the algorithms are far more refined.
Second, there is far less variability amongst humans than amongst birds (heron vs warbler) or amongst mammals (elephant vs french bulldog anyone?) making the algorithms far more complex for wildlife + pet detection.
Finally, with animals we are often fighting against evolution - they camouflage well in their environment - whereas humans, even when obstructed, have a talent to stand out.

Combine all 3 and I am not surprised that human detection in any brand is far more advanced but I also marvel at how far animal detection has come - and it will get better.
 
Having worked back in college with facial and object detection, it's been my experience that human faces are quite easy for a machine to pick up (once you apply the math required, we all kind of look the same to a machine). If you take into account that our skin colour is on limited intervals, you could do reliable face and eye detection 16 years ago on limited hardware.

When it comes to animals, the issue is with creating the general algorithm/neural network/etc that can cover the most use cases. Because each new data set you use to train it with will lead to performance droping for prevoius data sets (so you might end up with a brilliant system for detecting geese but it would be rubbish at anything else).

It might be interesting if any manufacturer would allow us to bias their animal detect system twards specific species (e.g: I know I'm gonna shoot bears so I'm gonna feed it a load of bear data sets to bias it twards these then reset it when I'm done with it).
 
Having worked back in college with facial and object detection, it's been my experience that human faces are quite easy for a machine to pick up (once you apply the math required, we all kind of look the same to a machine). If you take into account that our skin colour is on limited intervals, you could do reliable face and eye detection 16 years ago on limited hardware.

When it comes to animals, the issue is with creating the general algorithm/neural network/etc that can cover the most use cases. Because each new data set you use to train it with will lead to performance droping for prevoius data sets (so you might end up with a brilliant system for detecting geese but it would be rubbish at anything else).

It might be interesting if any manufacturer would allow us to bias their animal detect system twards specific species (e.g: I know I'm gonna shoot bears so I'm gonna feed it a load of bear data sets to bias it twards these then reset it when I'm done with it).
Extra points to the manufacturers that licenses Merlin from Cornell university and uses their algorithm + gps info to predict species and use that info to increase AF performance 😁
 
Agree with these comments.

What I am wondering about is if the (slight) deficit in af performance apparently seen with the Z9 versus the A1 is solely due to differences in eye detect performance for birds and or animals.

I mean, I just don't see problems with eye detect in humans regardless of clutter in the foreground or background. It just sucks the human eyeball out from all this. (Garish imagery for effect. )

Does anyone have any possible insight into this?
 
Agree with these comments.

What I am wondering about is if the (slight) deficit in af performance apparently seen with the Z9 versus the A1 is solely due to differences in eye detect performance for birds and or animals.

I mean, I just don't see problems with eye detect in humans regardless of clutter in the foreground or background. It just sucks the human eyeball out from all this. (Garish imagery for effect. )

Does anyone have any possible insight into this?
The only insight I have agrees with you - the Z9 is insanely good at sticking to human eyes. I haven't used it a ton in that arena, but man, every time I do I'm blown away. It's also good with most animals and birds. The only time it seems to struggle a bit is with fast moving targets. I'm using it in Africa and for most subjects it's easily finding and sticking to the eye. For general wildlife, it's doing a good job with eye detection and really only missing the tough, weird eyes (elephants, hippos).
 
@Steve :

If I remember correctly, the Z9 has an auto mode for subject identification as well as the ability to tell it the class of subject you are shooting.

Have you seen any performance difference between letting the camera figure out what it should detect and you telling it you are shooting birds/animals?
 
it takes time, sony has had a several year head start

Also from my college days, the algorithms used back then for subject detection (back then Fuji was big on this, having a lot of cameras with "fd" suffix) were developed by a firm called FotoNation. They were head-hunting in our college and one of their selling points was displaying their algorithms at work.

Chances are that both Nikon and Sony have licensed their algorithms from a 3rd party and Sony are either running them on slightly better hardware, had better training sets or they just paid for the better software :)
 
The only insight I have agrees with you - the Z9 is insanely good at sticking to human eyes. I haven't used it a ton in that arena, but man, every time I do I'm blown away. It's also good with most animals and birds. The only time it seems to struggle a bit is with fast moving targets. I'm using it in Africa and for most subjects it's easily finding and sticking to the eye. For general wildlife, it's doing a good job with eye detection and really only missing the tough, weird eyes (elephants, hippos).
I think I know the answer to this question, but I'll ask anyway. What about an elephant's or hippo's eyes is causing the camera to have issues?
 
@Steve :

If I remember correctly, the Z9 has an auto mode for subject identification as well as the ability to tell it the class of subject you are shooting.

Have you seen any performance difference between letting the camera figure out what it should detect and you telling it you are shooting birds/animals?
I also have some experience with recognition algorithms. My impression is that restricting the subjects, like to only humans, makes the Z9 quicker and more certain. Also restricting the area it needs to search. I haven't done any controls for this so there is a lot of confirmation bias here.

It is my prejudice, based on now outdated experience, is that the more possibilities you give a recognition algorithm, the more false positives you will get.

Conversely, if you don't train it on enough possibilities, it will give false negatives.

The Z9 also considers body, head and eyes, apparently in that order. I haven't heard that the others do this and I wonder if this overhead is holding it back. Presumably this approach is useful when the eyes are not visible but it surely adds computational overhead and potential for false detections. Sounds like a tradeoff.
 
The Z9 also considers body, head and eyes, apparently in that order. I haven't heard that the others do this and I wonder if this overhead is holding it back. Presumably this approach is useful when the eyes are not visible but it surely adds computational overhead and potential for false detections. Sounds like a tradeoff.
personally i don’t think so. the hierarchical selection of body part is a limited set and merely determines where on the body it focuses on, not on the ability to determine the subject itself.
 
I also have some experience with recognition algorithms. My impression is that restricting the subjects, like to only humans, makes the Z9 quicker and more certain. Also restricting the area it needs to search. I haven't done any controls for this so there is a lot of confirmation bias here.

It is my prejudice, based on now outdated experience, is that the more possibilities you give a recognition algorithm, the more false positives you will get.

Conversely, if you don't train it on enough possibilities, it will give false negatives.

The Z9 also considers body, head and eyes, apparently in that order. I haven't heard that the others do this and I wonder if this overhead is holding it back. Presumably this approach is useful when the eyes are not visible but it surely adds computational overhead and potential for false detections. Sounds like a tradeoff.
In my experience, contrary to most camera branding marketing of subject detection based on ai learning, I don’t think we can “train” to make the camera better recognise the subject. The “training” part I believe is done directly from the manufacturer. I do not feel any improvement in subject recognition over prelong use and most of the mirrorless camera experience change or improvement in subject recognition after firmware change.

Also I shoot people and animals, when I am on animal detect (specifically on z9) it will rarely pick up the human eye. This shows me that indeed it does have the ability to differentiate to a certain extend. this exclusive feature can be use to our advantage if use correctly.
 
In my experience, contrary to most camera branding marketing of subject detection based on ai learning, I don’t think we can “train” to make the camera better recognise the subject. The “training” part I believe is done directly from the manufacturer. I do not feel any improvement in subject recognition over prelong use and most of the mirrorless camera experience change or improvement in subject recognition after firmware change.

Also I shoot people and animals, when I am on animal detect (specifically on z9) it will rarely pick up the human eye. This shows me that indeed it does have the ability to differentiate to a certain extend. this exclusive feature can be use to our advantage if use correctly.
Seems to me that the more distinct objects it's trained for, the more you are required to expand feature space. Features, or more specificly sets of features, are the traits that distinguish one object from another.

I agree. The individual Z9's show no sign of learning...That all comes from the factory. I also would guess that Nikon, or any ca era other manufacturer, cannot do it all on their own.

There is a big industry that develops training data bases possessing the richness to cover what will be encountered in the wild. And there is the corresponding effort to determine the principle components of feature space that optimally distinguish one object from a other. It's a lot more involved than just feeding it a deck of Audubon Society flash cards.
 
I think I know the answer to this question, but I'll ask anyway. What about an elephant's or hippo's eyes is causing the camera to have issues?
Subject detection often seems to look for eyes-in-a-face and elephants and hippos have kind of weird-looking features. That, and elephant eyes at least are often covered with heavy eyelashes. It does seem to find giraffe eyes in profile really well :)
 
Its should not be be so hard is what i don't like about mirror less in the Z9, i think manufactures are trying to hard to be something for everyone in one camera.........they are unnecessarily complicated often at the expense of creativity and enjoyment.
Some of my best winning shots came from my D3s, D3X, DF. LOL.........its because of immersing myself in the environment and opportunity at hand...........

The best camera in the world is the one you have in your hand at the time.

Here is a candid snap from the D3s 12 mp on the 300 2.8 vr II so easy LOL



_DSC1760_4.jpg
You can only see EXIF info for this image if you are logged in.
 
I haven't read all this thread yet and got referred to it from following another thread. But I think this thread deals with specific equipment and technical ability ( possibly gender based ), but all I know is that Ansel Adams with his "antique" equipment produced artistic work that I never will be able to produce regardless of how advanced or expensive my equipment is or how much learning I am able to accomplish in my time left on this world. As for gender, many of our most technical, scientific, mathematical, medical and engineering leaders are female and as a male, I am proud to say that they are my peers and mentors teaching me.

Let's put to rest that equipment creates art and that one sex is inherently more capable than the other. That said, yes for photography you need the right equipment for the job, especially for capturing fast objects in poor lighting but how one uses the proper tools has no absolute right or wrong. I didn't do surgery using a chain saw, i used a scalpel under the brightest lighting I could have. Even though retired for 8 years, it bugs my wife that I have to turn on the overhead light just to wash the dishes or brush my teeth.

But i will admit to the writers of this thread that I am naturally drawn to the latest and newest technology and added the Z9 to my Nikon Collection of Z cameras specifically for learning about wildlife photography from my first love of Landscape and am waiting patiently for my Z 800 mm. But since having the Z9, I realized how much I missed the ergonomics of a full sized camera body ( like the D810 cf the Z 7ii ). Initially I thought I would use the Z7ii for my landscape and just use the Z9 for wildlife / nature with longer lenses, but now regardless of what I plan on shooting, I first pick up the Z9. It is just a joy to use even though my Z7ii packs lighter, takes up less space and takes just as great Landscape / candid etcetera images. Go figure !!

You guys and gals are the greatest and highly respectful of each other !!!
 
Its a question that has been beat to death on social media sites but I want to ask it in a different perhaps simpler way of all Nikon users. Its simple "would you trade in your D4S, D5 or D6 for a Z9"?
No I sold them and now have two Z9 and still own and am keeping two D850 (for now). I have added some big ticket Z-mount long glass and the f/2.8 zooms and will be selling off the f-mount version of these lenses soon. BUT not my large collection of shorter f-mount primes and pc-e lenses. I for one have no need for an f/1.2 50mm or any f/1.8 short prime.
The 45.7MP Z9 is a step change for Nikon shooters -- particularly those of us who own a lot of long Nikkor glass, use the same in trying environments and who have used the system for a long time -- many decades in my case. The D6/D5/D4S had the same 21mp sensor, which was a problem for me.
The Z9 is HOWEVER not a panacea -- no camera is. We learn the new body and adapt our use to minimise any, hopefully short term, deficiencies. AND we provide feedback to nikon via our NPS reps with the hope that the next firmware update or product release addresses the issues we have found -- see today's V1.20 firmware update for the 70-200 - this was almost entirely fixes.
Young eyes fingers and legs or decades of experience are what are needed to get great shots with any camera. The Z9 just makes it far far easier than it was before and every firmware update just makes it better. The 400/2.8TC and 800/6.3 PF give us better solutions and less weight.

I would like Nikon to come out with a Z9"S" whose base ISO is 200 not 64 - if such a change would give me the same or better low light performance/quality as my D6/D5 used to (before I sold them)

AND I wait to see what the sensor on the Z8 might be before I will part with my pair of D850 -- still the best Nikon sensor by far for still life/portraits etc... The Z7 is not - nor is the Z7II a D850 replacement - I doubt a Z7III would deliver a vastly better sensor than in the current version on top of adding an EXPEED 7 and other better/faster technology.

To those Brand Loyalists from Sony and Canon who cannot help making snarky comments about the Z9 -- wouldn't it be nice to have a camera that did not overheat while shooting vids, that is actually usable by people with large hands and ...... the list is equally long for any model you might own or use. So just stop it.
 
To those Brand Loyalists from Sony and Canon who cannot help making snarky comments about the Z9 -- wouldn't it be nice to have a camera that did not overheat while shooting vids, that is actually usable by people with large hands and ...... the list is equally long for any model you might own or use. So just stop it.

I have large hands and I prefer the size/shape of the a1. No Z9 snark, just a personal preference. Brand doesn't matter, I'll keep an eye on all the makers for a camera that combines the size & weight of the PF lenses with the size, weight & performance of the a1.
 
I have large hands and I prefer the size/shape of the a1. No Z9 snark, just a personal preference. Brand doesn't matter, I'll keep an eye on all the makers for a camera that combines the size & weight of the PF lenses with the size, weight & performance of the a1.
The size "issue" is easily solved with the A1. I use a small and light Meike grip extender to solve the floating pinky issue, and carry an extra battery so have zero battery life issues.
The size and especially weight issue (for me) with the Z9 though is not solvable, I can not cut its size nor its weight.

I can see the place for flagship type cameras, but as a non-pro have zero interest in them. I am on a waiting list for the Z800PF, but only with the leap of faith that Nikon will at one stage offer a non-flagship type action camera.
 
I have been waiting for a camera like the Z9 for years, since Nikon announce its mirrorless cameras and finally having it in my hand, i absolutely love it.

It is built like a tank, its autofocus is fast and precise, to me, it never misses a beat. It's an amazing camera for both video and stills and more importantly to me the colors are pleasant and the files are flexible and easy to work with. In my experience, its image quality is on par with the D850. But the autofocus and speed is in an entirely different league.

I tried Sony for a while (the A9II) and while i absolutely see why it attracts a lot of people, it just was not for me. I often found that the colors were off, it may be my old Nikon bias but it often took me a long time to correct skin tones. I liked the 200-600 zoom, it's nice versatile range but i found that the copy i had, had a hard time keeping up to my expectations in both sharpness and more importantly accuracy. It may have been a bad sample, or perhaps i just never got used to it, but it just never really worked out all that well for me. In the end i sold it.

It was not a bad experience though. Sony's autofocus system is very reliable and i still miss some of the Sony prime lenses for portrait shoots especially the 135mm 1.8G was really nice. I have never tried the A1 but i imagine that it's a fantastic camera and if you can afford the big Sony primes i think it's a winning combination. My own budget did simply not allow for those big Sony primes, and without them in my opinion, the system did not live up to my expectations for wildlife.

Today i have the Z9, a 24-70mm 2.8s, an old 58 1.4G for special looks, and a 85mm 1.8S for pin sharp portraits. I previously had the Sigma 500 F4 which worked amazingly on the Z9 for my wildlife needs but sold at the announcement of the new 800mm S lens and and i am now rather impatiently waiting for it to turn up.

My Z9 system with all the lenses, adapters, memory cards and lighting cost me about $17.000 (European prices converted) a similar Sony system around the A1 would be twice that with a 600mm F4. So both budget wise and user experience wise i am very happy.

The Z9 IS a really good camera.
 
Last edited:
Back
Top