John Navitsky
Well-known member
due to the size of the box we can deduce that subject detection is not “seeing” the eye. from that i would consider that perhaps it’s not good at swans or perhaps it was too far away
Questions regarding Nikon Z8 Eye AF - animals
- Thread starter ajrmd
- Start date
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).
- Status
Hood induced? This is an extremely interesting comment to me. For a long while I was observing what seemed to me to be the most unusual and questionable pattern - so much so that I wrote it off as me being "crazy" - that whenever I would say the heck with it and not use the lens hood on my 200-500 the photos would be much better in sharpness. I assumed I must be wrong because everything I have seen says to use a hood for best results in IQ - but you are saying the lens hood could produce a kind of heat distortion? I would greatly appreciate more details.
Cameron T
Well-known member
Yep. Heat gets trapped in the hood, and with no (or less) air movement the temperature difference from inside the hood to outside it can make images soft.
It's really circumstance dependant if you'll run into it, just like a lot of atmospherics.
SoCalRick
Well-known member
My apologies, as it looks like some of my earlier comments were incorrect. When I checked the Z 9 Online Reference Guide, it does indicate that the image should show the focus point used when the picture was taken:
Later on it states: If the camera can detect neither face nor eyes, it will display a focus point over the detected animal. This could be the reason the AF point is showing up on the subject's body and not their eyes in some of your examples.
Again, sorry for the earlier confusion.
Later on it states: If the camera can detect neither face nor eyes, it will display a focus point over the detected animal. This could be the reason the AF point is showing up on the subject's body and not their eyes in some of your examples.
Again, sorry for the earlier confusion.
Last edited:
John Navitsky
Well-known member
yes, unfortunately that is the nature of machine learning. it can be better at airplanes because 1) they are less varied and/or 2) it was trained on similar shaped airplanes.
and conversely it can have problems with a type of bird because 1) there are more shapes of birds than planes and 2) and/or it may not have been trained on similar shaped birds as much as you need in this particular case
John Navitsky
Well-known member
Steve covers it in one of his videos. This is often encountered shooting out of a window. Temp is likely warmer inside (but same problem as long as the temperature is different inside vs outside by enough), that warm air is sitting inside the hood, then you stick then lens out the window into the cooler air and you have heat distortion in a subtle and hard to debug way
John Navitsky
Well-known member
right, this is why i say we can infer what the camera is seeing by the SIZE of the af box
Palouse
Well-known member
That's what I guessed. Red square can't keep up.
Also covered in Steve’s book which I can’t access at the moment is two other factors: long necked birds are harder on the AF, and the order of acquisition: body, head, eye and then the reverse if it loses the eye may be what’s in play. My guess is that for a bird as far off as some of these images unless there is a ton of contrast the camera never found the eye.
Sometimes going into DX mode can help as the eye becomes larger.
Try shooting some short neck birds in DX mode and see what happens.
Or they similar birds after getting much closer.
Sometimes going into DX mode can help as the eye becomes larger.
Try shooting some short neck birds in DX mode and see what happens.
Or they similar birds after getting much closer.
JimC
New member
Here is my experience with my Z8 and Z9 using primarily my Z800mm lens:
I usually use focus area mode Wide S or W-C1 set at 1x1 with animal detect and 10 fps. Using herons and bald eagles as examples, the cameras "see" the head/eye quite easily whether in good light or mediocre light. I'll use the example of very good light monitoring a bald eagle some distance away in the top of a fir tree. The small focus square is right on the eye according to Nikon NX Studio software. When the eagle launches into the air, the focus point almost always falls off the head and frequently off the body entirely - the indicated focus point remains the small square as if it is still seeing the eye/head. Then often it catches back up to the eagle, showing on the body or close to the bird - still the small focus square ( when the camera drops focus on a stationary bird - say the head turns away - the focus square increases in size showing the body as the focus target). The focus square rarely catches back up to the head or eye. This all occurs while in 3D-tracking mode.
When it has occurred in Wide S or W-c1, I have assumed I am just not keeping the focus point in the proper location, but ????
The fact that the focus point indication remains the small square, as occurs when the camera moves from body to head to eye and many of my eagle photos look to be in very good focus, I wonder if the indicator of focus point in the camera is not keeping up with the transfer of digital data?
Now in the case of a heron flying towards me (usually off to the side a ways) and I have not gained focus on the eye or head prior to that point, I just don't see the eye picked up. The focus point is usually on the bird, often a wing, and sometimes I get it on the head - but not the small focus square. I have assumed this is due to me not getting a good initial lock or poor lighting. But if the heron was not flying in the same conditions, the camera would pick up the eye quickly.
I usually use focus area mode Wide S or W-C1 set at 1x1 with animal detect and 10 fps. Using herons and bald eagles as examples, the cameras "see" the head/eye quite easily whether in good light or mediocre light. I'll use the example of very good light monitoring a bald eagle some distance away in the top of a fir tree. The small focus square is right on the eye according to Nikon NX Studio software. When the eagle launches into the air, the focus point almost always falls off the head and frequently off the body entirely - the indicated focus point remains the small square as if it is still seeing the eye/head. Then often it catches back up to the eagle, showing on the body or close to the bird - still the small focus square ( when the camera drops focus on a stationary bird - say the head turns away - the focus square increases in size showing the body as the focus target). The focus square rarely catches back up to the head or eye. This all occurs while in 3D-tracking mode.
When it has occurred in Wide S or W-c1, I have assumed I am just not keeping the focus point in the proper location, but ????
The fact that the focus point indication remains the small square, as occurs when the camera moves from body to head to eye and many of my eagle photos look to be in very good focus, I wonder if the indicator of focus point in the camera is not keeping up with the transfer of digital data?
Now in the case of a heron flying towards me (usually off to the side a ways) and I have not gained focus on the eye or head prior to that point, I just don't see the eye picked up. The focus point is usually on the bird, often a wing, and sometimes I get it on the head - but not the small focus square. I have assumed this is due to me not getting a good initial lock or poor lighting. But if the heron was not flying in the same conditions, the camera would pick up the eye quickly.
With the caveat I mentioned earlier about hoping that a different lens may improve matters, I went through many of my old photos with NX studio (which I normally don't use) to look at the AF points and while I have lots and lots of little red boxes on birds' eyes and faces when they're walking around, I couldn't find one when they're flying.
Here's probably the most noteworthy example:
The photo is uncropped, so that bird is plenty large. It was moving mostly laterally but maybe a bit towards me. The head is mostly tucked in, so the long neck shouldn't be screwing things up too badly. I suppose the contrast could be better between the subject and the background, but it's not awful, especially in the area of the head. This was one of a burst of a good 20 shots or so and none of them were on the eye. Some of them were rejected during culling because the point was actually on the wing.
Here's another from the same sequence, the only other one from that sequence I kept:
Same bird, same day, but a considerably smaller size in the frame and possibly worse contrast with the background:
Another, different day:
Here's one of a tiny duck in the frame, terrible contrast with the background, and, you can't see it well here, but the eye is closed so it is the same color as the rest of the head:
Tiny, tiny bird, almost no contrast:
Just two more in flight:
Also, for whatever it's worth, unlike the OP I have a fair number of shots of airplanes and only one or two where it hit the cockpit: in post cases it focused right under the wing.
Here's probably the most noteworthy example:
You can only see EXIF info for this image if you are logged in.
The photo is uncropped, so that bird is plenty large. It was moving mostly laterally but maybe a bit towards me. The head is mostly tucked in, so the long neck shouldn't be screwing things up too badly. I suppose the contrast could be better between the subject and the background, but it's not awful, especially in the area of the head. This was one of a burst of a good 20 shots or so and none of them were on the eye. Some of them were rejected during culling because the point was actually on the wing.
Here's another from the same sequence, the only other one from that sequence I kept:
You can only see EXIF info for this image if you are logged in.
Same bird, same day, but a considerably smaller size in the frame and possibly worse contrast with the background:
You can only see EXIF info for this image if you are logged in.
Another, different day:
You can only see EXIF info for this image if you are logged in.
Here's one of a tiny duck in the frame, terrible contrast with the background, and, you can't see it well here, but the eye is closed so it is the same color as the rest of the head:
You can only see EXIF info for this image if you are logged in.
Tiny, tiny bird, almost no contrast:
You can only see EXIF info for this image if you are logged in.
Just two more in flight:
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.
Also, for whatever it's worth, unlike the OP I have a fair number of shots of airplanes and only one or two where it hit the cockpit: in post cases it focused right under the wing.
ajrmd
Well-known member
Your experiment confirmed my assertions. Thank you. The AF point registered in camera and in the software at the time the shutter is pressed is the one which appeared active in the EVF.
Interesting thread. My experience suggests that lenses can have an impact. I find that my 500pf is faster and more consistent at picking up the eye than my 100-400 Z lens. One might say that has to do with increased focal length but I find it true with closer BIF such as hummingbirds.
In my part of the world you can’t get closer than 100 feet to a GBH before it takes off and flies away from you. Rarely ever can you get the focus point on the eye. However in a recent trip to Pensacola I found that I could get much closer to GBH and the focus point nearly always got the eye when closer. That was with both the 500pf and the 100-400 s lens. I have also seen the same thing with sandhill cranes here in West Texas vs going to Bosque Del Apache.
I have my movie record button programmed to toggle between fox and DX on both my z8 and z9. It is a great tool in assisting to get better af on distant subjects. It helps a great deal when shooting BIF.
My experience tells me that the farther away from a BIF the less likely you will get the eye as the focal point. Take the same distance on a static subject and you have a greater chance of getting the eye as the focus point. Finally, I have never seen the part about long necked birds being more difficult to get the eye as the focus point but it makes a lot of sense from my experience.
I don’t have the 800pf, but with greater focal length I would assume thar it would be even more difficult to get the AF on the eye because of greater distance and possible atmospheric conditions.
In my part of the world you can’t get closer than 100 feet to a GBH before it takes off and flies away from you. Rarely ever can you get the focus point on the eye. However in a recent trip to Pensacola I found that I could get much closer to GBH and the focus point nearly always got the eye when closer. That was with both the 500pf and the 100-400 s lens. I have also seen the same thing with sandhill cranes here in West Texas vs going to Bosque Del Apache.
I have my movie record button programmed to toggle between fox and DX on both my z8 and z9. It is a great tool in assisting to get better af on distant subjects. It helps a great deal when shooting BIF.
My experience tells me that the farther away from a BIF the less likely you will get the eye as the focal point. Take the same distance on a static subject and you have a greater chance of getting the eye as the focus point. Finally, I have never seen the part about long necked birds being more difficult to get the eye as the focus point but it makes a lot of sense from my experience.
I don’t have the 800pf, but with greater focal length I would assume thar it would be even more difficult to get the AF on the eye because of greater distance and possible atmospheric conditions.
ajrmd
Well-known member
"With the caveat I mentioned earlier about hoping that a different lens may improve matters, I went through many of my old photos with NX studio (which I normally don't use) to look at the AF points and while I have lots and lots of little red boxes on birds' eyes and faces when they're walking around, I couldn't find one when they're flying."
HALLELUIA! Finally, someone has posted what I asked for and has confirmed that I am not crazy. The Z8 can eye detect static or slowly moving/swimming birds though it appears unable to eye detect for BIF. Others have dismissed my observations and have provided rationales including subject distance, lack of contrast, atmospherics, etc. I've provided numerous examples of similar subjects under similar lighting at similar distance, under similar conditions where the camera easily documents eye detect on the static or slowly moving/swimming bird but does not appear to eye detect/track for BIF. Additionally, I've provided examples of images crafted with a competing system which demonstrates eye detect/tracking for BIF. As I have observed, the Z8 is capable of eye tracking for people walking, windshields for cars, and cockpits for fast moving jets/monoplanes. Again, I challenge users to post images with evidence of eye detect/tracking of BIF.
So, this raises several important questions. First is this an issue with Nikon's eye detect algorithms, machine learning, lens interface, internal processing, or other challenges? Additionally, how does this impact AF accuracy or "keeper rate"? My own Z8 portfolio contains several thousand static and moving images, which are not sufficient data compared to the hundreds of thousands of images shot with competing systems. The debate continues.
ajrmd
Well-known member
Please post a single BIF image with eye detect/tracking in action. The same tired rationales are getting boring. I've demonstrated that the camera is capable of focusing on the eye, tracking when the bird is perched, static, or swimming. At similar distances, conditions, etc. the system appears incapable of eye detect/tracking when the bird is flying. Trust me, I would rather be proved wrong with the capital investment I have made.
Last edited:
lwan
Well-known member
Is stabilization enbled?
If yes, then at the moment of capture the shift induced by the stabilisation is simply not taken into account in repositioning the af point, as it happened with canon and sony cameras i had in the past. I never used nx to see my actual recorded focus point with my nikon camera, but might be worth checking.
If yes, then at the moment of capture the shift induced by the stabilisation is simply not taken into account in repositioning the af point, as it happened with canon and sony cameras i had in the past. I never used nx to see my actual recorded focus point with my nikon camera, but might be worth checking.
ajrmd
Well-known member
Wrong. Eye detect works and records accurately on walking people, swimming birds, etc. with stabilization active. Cockpit detect works on fast moving airplanes and cars with stabilization active, so why doesn't eye detect/tracking appear on BIF? There's a disconnect. The camera can detect and track a cockpit of a jet moving hundreds of miles an hour and it can't detect/track an eye of a large bird flying at tens of miles an hour? Would someone post some images of say dogs or a mammal other than a human running with eye detect tracking either working or not?
Cameron T
Well-known member
Do you really expect people to potentially go through tens of thousands of photos each to prove you wrong? I know I'm not, and I know I have some photos somewhere where it found the eye in flight.
That's not exactly right. The camera finds a subject faster with a smaller AF area - and a smaller AF box almost always improves AF performance. Both Nikon Ambassador Reed Hoffman and Nikon specialist Paul van Allen made this suggestion in Z9 programs. It CAN find targets outside the AF box, but the box and the area immediately adjacent to the box is where the camera looks for the target. The first priority with subject recognition is to identify the right subject. So if the AF box is on a bird, it looks for a bird's head and then a bird eye in the box or near the box because it has identified the subject.. If it does not find a bird in the area with a box, it looks for a subject elsewhere in the frame. What the camera won't do is forego a subject under the box for one outside the box unless it is very near the box edge. Once the camera has found the subject, it progressively looks to the head and then the eye - but part of that is based on the relative size of the subject in the frame.
There are subject specific differences. The most recent Z9 firmware improved the distance for human eye recognition.
Also the boxes are not directly linked to focusing - they are simply a display. The boxes are a separate overlay and focus operates independent of the box. So there is some variation between actual focus and the appearance of an AF box.
I'll respond more later when I have time, but quickly I want to be clear to reiterate that I have absolutely seen examples of BiF eye focus in videos and I can't say the issue isn't more my lens.
ajrmd
Well-known member
Having made a significant capital investment in Nikon and wanting to capture better images, it is important to understand the capabilities/limitations of the system. If, for example the system is incapable of capturing a large bird in focus landing 20-30' away from me, I need to know this. If the system is incapable of tracking the eye of a BIF at whatever incident angle from the user, I need to know this.
What I have determined is that the Z8 is capable of eye detect and tracking for static people, birds, animals. Moreover, it appears to be able to eye detect and track people, birds, animals walking, swimming, or moving. It has very good recognition of cockpits for planes flying at hundreds of MPH and cars moving at high speed. What I would like to understand is whether this system is capable of eye detect and tracking for BIF?
What I have determined is that the Z8 is capable of eye detect and tracking for static people, birds, animals. Moreover, it appears to be able to eye detect and track people, birds, animals walking, swimming, or moving. It has very good recognition of cockpits for planes flying at hundreds of MPH and cars moving at high speed. What I would like to understand is whether this system is capable of eye detect and tracking for BIF?
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.
You can only see EXIF info for this image if you are logged in.
Again, there are videos which clearly demonstrate it is possible. I have even found a few in my old collection that I may post later when I'm back home. Also, you're basing all if this on a single copy if a single lens. I can understand being frustrated, as I have been at times due to a similar issue - BUT at the same time multiple people have explained that it's regularly worked for them or others they have seen and given that vs the experience of one or two people with limited lenses available, I don't think it makes sense to keep questioning the entire system.
Yes it is; and I'll say it again that in your images you appear to be too far away for it work reliably on the eye. Maybe buy Steve's book on Secrets to the Nikon AF system. Here are two sections on using DX to get closer and the process of Body>head>eye. He describes how he captures that sequence (although for a stationary bird) and the images are in the book. Still applies to moving birds. Next time you are out try to see if you can get eyes locked on at a particular distance and then try to catch them in flight at a similar distance.
On page 178 it says -
Further on page 167 he states:
On page 178 it says -
As a side note, one nifty trick with distant subjects where the camera
struggles to identify the face or eye is switching to DX crop mode. This
enlarges the image in the viewfinder - and in my testing the camera
sometimes picks up the face (or eye) just a bit easier. I mean, if you have a
distant subject you’re gonna crop anyway, why not do it in the field and
give subject detection a hand?
Further on page 167 he states:
The Z9 can detect the body, face, and eyes - and does so in that order. It
happens fast enough that it often seems like it’s just going for the eye, but
according to Nikon the sequence is body > head > eye. In fact, I’ve often
watched the system refine its AF position by starting on the body, finding
the face, and finally drawing a box around the eye. If you shoot the camera
long enough, you’ll certainly witness this yourself. I was lucky enough to
capture this as shown in the series of screenshots below.
JimC
New member
JimC
New member
These screenshots are from my Z9, Z800 PF lens with Focus Mode: AF-C and AF-Area Mode: 3D-tracking. The focus square was right on the eye while the Bald Eagle was essentially stationary, but drops the eye when the bird launches and is in flight. Focus overall seems good.
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.
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.
- Status