Questions regarding Nikon Z8 Eye AF - animals

[ajrmd]

As a recent Nikon convert (coming from Canon/Sony), I've been spending some time trying to better understand the AF system (configuration/operation) and have read/watched nearly every tutorial. Now that I have several thousand images, I've taken some time to review what is working and what is not and trying to troubleshoot where I am falling short. Overall, I've been very pleased with the AF performance for most static subjects. The eye detect and tracking works very well for humans and most animals (could improve on ungulates but that's true of every AF system). Also, the detect and tracking is superb for airplanes and vehicles, and it does nail the cockpit nearly every time! Now, turning to BIF, I am noticing some challenges, and my observations correspond to the number of "in focus" captures. Before we explore this further, I understand that if one is shooting at extreme distance, the AF system is more likely to detect and track some point of the moving object; a wing, side, etc. and that is true of nearly every camera's AF system. As I review the active AF points in NX studio, I am noticing a disconnect between what I am viewing in the EVF and what the camera is actually capturing.

The following few images are examples from a longer series of a Sandhill Crane coming in for a landing. I had first engaged the subject with sufficient time for the AF system to detect and track the subject and the white box in the EVF was clearly on the eye, though the resultant AF point was not and all of the images are OOF (they are FF and not cropped). I appreciate that when the subject is smaller in the frame (as demonstrated in the swan image) the Z8 may have some difficulty identifying the eye so it grabs the nearest surface (in this case the body) and the image will be relatively sharp because of the DOF and the eye is in a similar plane. As I mentioned, this is representative of the kinds of misses that I am encountering quite frequently and these shots would have been easily captured with my Canon/Sony gear. FWIW, I have my camera set up mirroring Steve's suggestions. Thoughts?

You can only see EXIF info for this image if you are logged in.

You can only see EXIF info for this image if you are logged in.

You can only see EXIF info for this image if you are logged in.

You can only see EXIF info for this image if you are logged in.

 

[SCoombs]

What's the lens?

 

[SoCalRick]

It would be helpful to have more shooting info and/or EXIF data. What speed, aperture, focus modes, VR, etc. were you using?

Regarding the location of the AF point shown during playback, in Steve's Z8/Z9 eBook he says not to put too much weight into where the AF point shows up in the image (due to many variables). He says it’s possible that the camera doesn’t always record the exact point properly, especially if you’re using something like Auto or 3D.

In other words, we can't assume that was where the focal point/plane was located at the time of capture.

 

[EricBowles]

What AF Area mode are you using? The smaller the AF area, the better AF is - especially when the subject is small in the frame like the two swans. 3D or area mode for initial acquisition normally works, but when it doesn't I fall back to a smaller area using a Fn button. Sandhill cranes are pretty easy to follow using Wide modes if Area or 3D are not working.

The sandhill crane looks a bit underexposed. That can be a problem, especially when the eye or silhouette is not obvious.

 

[ajrmd]

It would be helpful to have more shooting info and/or EXIF data. What speed, aperture, focus modes, VR, etc. were you using?

Regarding the location of the AF point shown during playback, in Steve's Z8/Z9 eBook he says not put too much weight into where the AF point shows up in the image (due to many variables). He says it’s possible that the camera doesn’t always record the exact point properly, especially if you’re using something like Auto or 3D.

In other words, we can't assume that was where the focal point/plane was located at the time of capture.

I anticipated a lot of questions would be asked about settings. These are screen captures from NX studio indicating the AF points. Are they accurate? Who's to say, though every in focus image, has the little box squarely over the eye for perched birds, people, planes, etc. I'm just not seeing this with my BIF images and am not changing any of the af settings. Defaults are as Steve describes and usually I acquire with a Wide C1 array, sometimes handoff to 3D or sometimes just acquire with 3d if it recognizes the subject with the little white square tracking box. In terms of the image settings for the examples provided, ISO 2500, Tv 1/2000, f/6.3 AF settings, AF-C, Autofocus area mode wide C1, Autofocus detection - animals A1 - release, A3 - 3, steady, A4 - all points used, A6 - AF, A7- Auto, A14 - High, etc. 800 PF.

 

[ajrmd]

What AF Area mode are you using? The smaller the AF area, the better AF is - especially when the subject is small in the frame like the two swans. 3D or area mode for initial acquisition normally works, but when it doesn't I fall back to a smaller area using a Fn button. Sandhill cranes are pretty easy to follow using Wide modes if Area or 3D are not working.

The sandhill crane looks a bit underexposed. That can be a problem, especially when the eye or silhouette is not obvious.

See above. Let's take a look at another image from a series of a slow moving GBH. Again, fairly large in the frame, exposure is ok, and it is unable to recognize the eye and instead focuses on the wing/body. Due to DOF, it is sufficiently large enough to be in focus though of the dozen odd shots all of the AF points were on similar places. Contrast that with a static shot of a cygnet swimming at a slightly further distance away several minutes later, again the AF is on the eye. I could post thousands of images of birds, mammals perched, slowly walking or swimming and nearly 100% of the time the AF point is on the eye. Not so for BIF even at reasonable and similar distances. Again, I am really trying to understand what is happening here and why.

You can only see EXIF info for this image if you are logged in.

You can only see EXIF info for this image if you are logged in.

 

[John Navitsky]

These are screen captures from NX studio indicating the AF points. Are they accurate?

i take the overlays with a grain of salt. my personal read on it is if the box is eye sized then it probably is tracking the eye even if the box isn’t shown in the right place. one has to remember of course that af and subject detection are two different things and there is a difference between deciding the right thing to focus on and achieving af. do find the setting that changes the box color when it thinks af is successful. this is helpful feedback

 

[ajrmd]

i take the overlays with a grain of salt. my personal read on it is if the box is eye sized then it probably is tracking the eye even if the box isn’t shown in the right place. one has to remember of course that af and subject detection are two different things and there is a difference between deciding the right thing to focus on and achieving af. do find the setting that changes the box color when it thinks af is successful. this is helpful feedback

I would like to agree, however the AF points do correlate with the in-focus images and as I have articulated, nearly 100% of the slowly moving or static people/animals have an AF point over the eye whereas none of the BIF do. Likewise, looking at my vehicle/flight images, the AF points are located over the cockpit. Here is one example and all of the images in this pass have the AF point over the cockpit... There is something happening with my BIF images and I am trying to troubleshoot it.

You can only see EXIF info for this image if you are logged in.

 

[EricBowles]

See above. Let's take a look at another image from a series of a slow moving GBH. Again, fairly large in the frame, exposure is ok, and it is unable to recognize the eye and instead focuses on the wing/body. Due to DOF, it is sufficiently large enough to be in focus though of the dozen odd shots all of the AF points were on similar places. Contrast that with a static shot of a cygnet swimming at a slightly further distance away several minutes later, again the AF is on the eye. I could post thousands of images of birds, mammals perched, slowly walking or swimming and nearly 100% of the time the AF point is on the eye. Not so for BIF even at reasonable and similar distances. Again, I am really trying to understand what is happening here and why.

My take is the GBH is still small in the frame for the eye to be relevant. As you said - DOF has the entire bird in focus anyway. The size of the eye and subject is relevant. The camera uses a progression - recognizable shape, body, head, then eye. The smaller the AF area being used, the more concentrated the camera is on a specific location. If it is on a subject and staying on the subject, it continues to refine and maintain focus.

In the case of the cygnet, you have the best possible situation - a dark eye which is well defined against the white head, and as much time as needed to determine that is the eye on the head of a bird. Compared to the GBH, the heron has a two tone face, the eye has low contrast in a busy patterned area, and the camera correctly found the bird but not the eye. I'd be using Wide Small or C1 for that kind of subject and if positioned on the neck, head, or shoulder.

You may find 3D works effectively for any of these subjects. If you have a relatively large subject, it works easily. If your subject is smaller in the frame, a smaller AF area such as Wide Small or Wide C1 is more effective to acquire focus, and then 3D can maintain AF and refine as needed.

To accomplish the latter, I use a Fn button programmed for Wide Small + AF-On - so the camera has a small box, and when you press the button it changes to Wide Small and immediately focuses. The camera will look for the head/eye in and near the Wide Small box. Once acquired, you can release the Fn button and return to your previously selected AF Area mode - 3D - and it should maintain the same focus on the subject and continue to adjust as you follow the subject.

Of course, you could also take other approaches. Some people use Wide Small as an example as the Primary AF mode, and find the subject. Once they find it, they press a button for 3D and 3D takes over as long as the button is held. It's a matter of personal preference.

I'm sure some others will weigh in with their suggested approaches. It's a bit of a personal choice because there are a lot of ways to configure the camera/

 

[ajrmd]

My take is the GBH is still small in the frame for the eye to be relevant. As you said - DOF has the entire bird in focus anyway. The size of the eye and subject is relevant. The camera uses a progression - recognizable shape, body, head, then eye. The smaller the AF area being used, the more concentrated the camera is on a specific location. If it is on a subject and staying on the subject, it continues to refine and maintain focus.

In the case of the cygnet, you have the best possible situation - a dark eye which is well defined against the white head, and as much time as needed to determine that is the eye on the head of a bird. Compared to the GBH, the heron has a two tone face, the eye has low contrast in a busy patterned area, and the camera correctly found the bird but not the eye. I'd be using Wide Small or C1 for that kind of subject and if positioned on the neck, head, or shoulder.

You may find 3D works effectively for any of these subjects. If you have a relatively large subject, it works easily. If your subject is smaller in the frame, a smaller AF area such as Wide Small or Wide C1 is more effective to acquire focus, and then 3D can maintain AF and refine as needed.

To accomplish the latter, I use a Fn button programmed for Wide Small + AF-On - so the camera has a small box, and when you press the button it changes to Wide Small and immediately focuses. The camera will look for the head/eye in and near the Wide Small box. Once acquired, you can release the Fn button and return to your previously selected AF Area mode - 3D - and it should maintain the same focus on the subject and continue to adjust as you follow the subject.

Of course, you could also take other approaches. Some people use Wide Small as an example as the Primary AF mode, and find the subject. Once they find it, they press a button for 3D and 3D takes over as long as the button is held. It's a matter of personal preference.

I'm sure some others will weigh in with their suggested approaches. It's a bit of a personal choice because there are a lot of ways to configure the camera/

So, in this case you rationalize that the cygnet has a dark eye against a white head and that allows for eye focus, yet in every series of swans flying, even those which fill the frame it is unable to find the eye? Again, the camera can identify a cockpit on a plane from a great distance, a car windshield along a track, a person's eye when walking, a bird's eye when swimming or perched, but it can't when the bird is flying? I'm confused.

 

[ajrmd]

100% of the nearly 40 images in this series of a swan in flight has the AF point somewhere on the body or wing. None on the eye.

You can only see EXIF info for this image if you are logged in.

 

[Cameron T]

I'll agree with the point about the box being potentially wrong, and just ignoring it. While it can be shown correctly, it's not a guarantee, and I never even bother with NX studio, because I just care if the shot was in focus or not. Not where an arbitrary box was recorded.

As for AF issues, I can't say I had them, but there seems to be times when one person has issues that nobody else can replicate, and it could be the firmware didn't take correctly, etc. I'd suggest handing your camera off to someone else for a bit and see if their shots lead to the same problems or not. If not, then the problem isn't the camera.

 

[SoCalRick]

For understanding what is going on in your situation, I believe that it is more important to know what AF information the camera is providing you in the EVF at the time of capture, rather than trying to read too much into the representation of the AF point on the image afterwords. Since the AF subject detection works by trying to detect a body, then a head, and then the eye, you should be able to see confirmation of that in the EVF. If the camera has locked onto the eye (and even as it progresses through the body, head, eye sequence), then it will show you confirmation of that in the EVF when you are in the field. When you see eye lock-on has been confirmed, then take your shots at 20 fps and pixel peep the results afterwords to see if you are getting sharp eyes, or not (independent of where the AF point is rendered on the image). If you are not getting sharp eyes, then what part of the image is sharp (if any)?

In your BIF examples you provided are you able to identify an area of the image that is in focus, or is nothing particularly sharp?

In some of the other examples that you mention, I think you may have too many variables changing, in order to make a valid conclusion about the results that you are getting. E.g., I'm guessing you are probably using different subject detection settings (e.g., People, Animal, Vehicle, Auto, None) in the examples you gave? You may also be using different AF modes for the different shots. And 3D tracking is also probably it's own unique variable by itself. Were you in 3D tracking for all the captures, some of them, or none? Were you using the same lens in all instances? Distance to subject, size in the viewfinder, etc.

I'm sure you've thought about limiting the number of variables that change when you try to replicate your issue and compare your BIF results with other shots, but you need to narrow the variables down so that you are making an apples to apples comparision (as much as possible).

Also, do you have another wildlife lens that you can use as a comparison to see if you can get better results for sharpness? The 800 PF is not the fastest lens out there, but it's also no slouch either, and it should function adequately at the ISO 2500 setting that you mentioned above.

Just my $0.02

 

[ajrmd]

I am curious to hear whether any users who shoot BIF have an image with an eye detect and not somewhere else on the body? Again, I've poured through thousands of images and am unable to find a single BIF where there is an eye detect (as evidenced by a small square over the subject's eye when viewed in post in Nikon NXstudio) at distances where eye detect should be active, i.e. had the subject was static, it would demonstrate eye detect. As an experiment, I just shot about 200 images of a person walking or static, a mammal or bird perched in wide area C1 or 3d and every single image had an eye detect in the evf and it corresponded in NX studio. Can someone post an image of a BIF with the eye detect active and not an AF point somewhere else on the bird's body, wing, etc.?

 

[Palouse]

What frame rate were you shooting at? Are all the images in a sequence similarly not covered by the red focus point rectangle?

 

[ajrmd]

What frame rate were you shooting at? Are all the images in a sequence similarly not covered by the red focus point rectangle?

20 FPS, RAW, and answer is it depends. So take the swan sequence with 40 images, all of them have a red focus point rectangle though it's on various places, neck, wing, back, etc. Sometimes, as in the first three images, it may or may not be.

 

[ajrmd]

So, here are two images with my R5 at similar FL's with the AF points clearly on the BIF's eye (or thereabouts). These are also uncropped and are pretty comparable in terms of subject size and type with the Z8 images I've posted. Does the R5 use eye detect all of the time? Nope, sometimes it focuses on the lead wing when the bird is flying towards the photographer when the eye is not visible, at other times it might be on the body as in the pair of swans in a prior post. I can't find a single Z8 image of mine where it appears to eye detect for a BIF. Can someone post a couple examples?

You can only see EXIF info for this image if you are logged in.

You can only see EXIF info for this image if you are logged in.

 

[flashpixx]

My experience is similar to yours where 3D tracking is used, both Z8 and Z9.
There is no way to control where the AF point is using 3D - the camera does it for you, so getting the eye in focus, even when using animal subject detection, is determined by the camera.
So, the closer the bird is and the bigger it is in the VF, the more likely 3D with subject detection enabled, the more chance of getting the eye in focus.

 

[SCoombs]

I have also reported pretty strong success in getting the focus on the eye during a static or slow moving shot but never having seen it get on the eye during flight, even when the bird is, say, the size of the DX frame. I've taken to assuming it's due to either the slow AF or some other aspect of my 200-500 F mount lens and have been hoping to see better results when I upgrade to the 180-600 because lots of people, like we see here, have reported that they get good bird eye focus in flight, and I've seen it in many videos, too.

On the one hand, I have previously seen much improved rates of hitting the eye for people when switching from an adapted F mount lens to a new Z mount lens of similar focal length. On the other, I admit to being skeptical I'll see much difference and don't know how to reconcile that with all the other reports people make of how it works well for them. I'd be open to blaming some aspect of my technique, but really what technique is there more than keeping the bird in the frame?

 

[ajrmd]

My experience is similar to yours where 3D tracking is used, both Z8 and Z9.
There is no way to control where the AF point is using 3D - the camera does it for you, so getting the eye in focus, even when using animal subject detection, is determined by the camera.
So, the closer the bird is and the bigger it is in the VF, the more likely 3D with subject detection enabled, the more chance of getting the eye in focus.

OK, well this occurs with 3d or wide array. Can you post a Z8 or Z9 image where the eye detect is clearly on the eye? Quite candidly, every sharp BIF Z8 image I have taken the camera focused on an adjacent part and either the eye is in plane or there is sufficient DOF to make it "in focus". Can the Z8 or Z9 utilize eye detect for BIF images?

 

[SCoombs]

My take is the GBH is still small in the frame for the eye to be relevant. As you said - DOF has the entire bird in focus anyway. The size of the eye and subject is relevant. The camera uses a progression - recognizable shape, body, head, then eye. The smaller the AF area being used, the more concentrated the camera is on a specific location. If it is on a subject and staying on the subject, it continues to refine and maintain focus.

In the case of the cygnet, you have the best possible situation - a dark eye which is well defined against the white head, and as much time as needed to determine that is the eye on the head of a bird. Compared to the GBH, the heron has a two tone face, the eye has low contrast in a busy patterned area, and the camera correctly found the bird but not the eye. I'd be using Wide Small or C1 for that kind of subject and if positioned on the neck, head, or shoulder.

You may find 3D works effectively for any of these subjects. If you have a relatively large subject, it works easily. If your subject is smaller in the frame, a smaller AF area such as Wide Small or Wide C1 is more effective to acquire focus, and then 3D can maintain AF and refine as needed.

To accomplish the latter, I use a Fn button programmed for Wide Small + AF-On - so the camera has a small box, and when you press the button it changes to Wide Small and immediately focuses. The camera will look for the head/eye in and near the Wide Small box. Once acquired, you can release the Fn button and return to your previously selected AF Area mode - 3D - and it should maintain the same focus on the subject and continue to adjust as you follow the subject.

Of course, you could also take other approaches. Some people use Wide Small as an example as the Primary AF mode, and find the subject. Once they find it, they press a button for 3D and 3D takes over as long as the button is held. It's a matter of personal preference.

I'm sure some others will weigh in with their suggested approaches. It's a bit of a personal choice because there are a lot of ways to configure the camera/

I actually have to offer some rare disagreement with you here.

First, the size of the AF box does not impact performance of subject detection in the Z8/9. The subject detection works on anything in the frame and the AF boxes only serve to "confirm" for the camera that you want to focus on a subject it's already detected. This is per Thom Hogan, several notable wildlife figures, and really just experience, and it's easy to verify. Go find a person, or a dog, or a bird, or any other detectable creature and put the wide area small box on its rear end - the camera will still focus on the head/eye if it sees it. Move the box off the critter and it will stop focusing on yhe eye. Now move the box back to cover any part of it - a wing, an arm, a stomach, etc.- and it will grab the head/eye if it can. It will work the same with the 1x1 custom area, so it you put that so that the only thing in the box is on a person's thumb it will still focus on the eye if it sees it. On the other hand, changing from FX to DX mode does improve subject detection as it decreases the frame size.

That is, to me, pretty widely acknowledged so I've no doubt of it. A second point is one I admit I'm basing more on my experience only, which is that 3D tracking does not work any differently or better or with more stickiness than subject detection - at least not if subject detection is active. If one toggles it off, 3D tracking may work better- I haven't tried it. With subject detection on, I've not found it to be at all distinguishable from how subject detection works.

 

[SoCalRick]

The location of the AF point at time of capture, and whether the camera is actually focusing on the subject's eye or not (due to subject detection being on) are two different things.

For example, if you are using Steve's recommendation for Wide-area AF C1 set to a 1x1 square, with subject detection (animal) enabled, then if you place your AF point somewhere on the animal's body (and the camera recognizes it as an animal), then the Z8 and/or Z9 will typically focus on the subject's head or eye, even though that area lies outside of your 1x1 box. Now, when you review the picture that you have taken, what is the correct location for the AF point to be shown? It was placed on the subject's body and not the eye when the picture was taken. The important point is whether the camera focused on the head or eye at the time of capture. That's why I asked earlier what the camera was showing you in the viewfinder at capture, and whether there was an indication that subject detection had locked onto the head, or eye.

We can only infer what logic/algorithm Nikon is using for determing the placement of the AF point on the images that have been taken, but that doesn't mean it is incorrect if it is not showing it as being on the subject's head/eye.

I considered deleting this post, but I decided not to because it was included in further discussions. However, it appears that I have made some incorrect statements here, and I have revised my comments in a later post (#29).

 

[ajrmd]

The location of the AF point at time of capture, and whether the camera is actually focusing on the subject's eye or not (due to subject detection being on) are two different things.

For example, if you are using Steve's recommendation for Wide-area AF C1 set to a 1x1 square, with subject detection (animal) enabled, then if you place your AF point somewhere on the animal's body (and the camera recognizes it as an animal), then the Z8 and/or Z9 will typically focus on the subject's head or eye, even though that area lies outside of your 1x1 box. Now, when you review the picture that you have taken, what is the correct location for the AF point to be shown? It wasn't placed on the subject's eye when the picture was taken. The important point is whether the camera focused on the head or eye at the time of capture. That's why I asked earlier what the camera was showing you in the viewfinder at capture, and whether there was an indication that subject detection had locked onto the head, or eye.

We can only infer what logic Nikon is using in determing the location for the AF point on the images that have been taken, but that doesn't mean it is incorrect if it is not showing it as being on the subject's head/eye. <-- and this also doesn't mean that the camera didn't focus on the head/eye at the time of capture.

My understanding, contrary to your assertion is that the af point registered and displayed in the software is the active af point(s) at time of capture and not the initiating point. This is backed up by my observations, where if one uses a small array, points it at a static subject and the white box jumps to the eye outside the array, the actual af point recorded is the one over the eye. I just confirmed this. Again, open your BIF images in the Nikon software, view the af point and see if any are on the eyes indicating eye detect was active. Please post a screenshot here.

 

[SCoombs]

My understanding, contrary to your assertion is that the af point registered and displayed in the software is the active af point(s) at time of capture and not the initiating point. This is backed up by my observations, where if one uses a small array, points it at a static subject and the white box jumps to the eye outside the array, the actual af point recorded is the one over the eye. I just confirmed this. Again, open your BIF images in the Nikon software, view the af point and see if any are on the eyes indicating eye detect was active. Please post a screenshot here.

Rick Raises an interesting question which I'd never considered before. When the AF point is displayed on a photo, is the point that was recorded where the AF box was, or is it where the camera was actually (at least trying to be) focused?

I did a quick experiment. This first image was taken of the read LCD at the exact instant I pressed the shutter. I pressed the camera shutter with my right hand and the phone with my left at the same instant.

You can only see EXIF info for this image if you are logged in.

The following image is what the camera displayed when reviewing the photo on the LCD:

You can only see EXIF info for this image if you are logged in.

Here is what NX studio displays:

You can only see EXIF info for this image if you are logged in.

Here is a processed version, followed by a crop to help, given the lower forum resolution, show that the eye is actually in focus:

You can only see EXIF info for this image if you are logged in.

You can only see EXIF info for this image if you are logged in.

It would seem, then, that if subject detection is active that the AF point encoded in the file is actually the subject detection point and not simply the AF box.

(As an aside, this exercise also reminded me of how woefully inferior LrC's color rendering is of the Z8 files than NX studio and makes me more frustrated that I can't seem to get it to reproduce them the same way.)

 

[John Navitsky]

remember to go back to basics. find the focal plane by seeing what is most in focus in the frame. of course that doesn’t mean it’s what it was trying to focus on, but it’s your most solid data point.

there are a lot of variables unfortunately. we know it uses prediction based on the subjects velocity and any change in the velocity or direction can throw focus off enough to not be a keeper

we know atmospheric conditions can throw off image quality enough for it not to be a keeper

heat difference, including hood induced

subjects moving perpendicular to you should be moving very little relative to you and be relatively easy for the camera to keep in focus

different parts of the subject that are on the same or almost same plane should also both be in focus or near.

look at each image and think about what you know and try to come up with hypothesis about what happened and postulate what you might try next time to prove or disprove

i know that might not seem helpful, but that’s how i learned to shoot dogs in flight