ajrmd
Well-known member
Yes, exiftool has focus distance and a whole lot more, probably more than you care to know.
Nikon, better subject detection, please!
- 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
I don't think it's very accurate to be honest. At the very least it's accurate sometimes but you can never tell when those sometimes are.I had no idea that it's that simple. Thank you!
I discovered this when trying to so some general AF comparisons and went on to test it pretty carefully on my Z8 uaing the following lenses:
180-600
500pf
70-200s
70-180
24-120s
85/1.8s
40/2
To test this, I chose a street sign that has good vertical contrast. I then physically measured the distance from the sign to three different locations. At each location, I took 3 photos with each lens. For zoom lenses I took three photos at multiple focal lengths and I was sure to do focal lengths shared between the lenses. For example, on my 70-200 and 24-120 I shot photos at 85mm because I was also testing the 85mm lens. On the 180-600 I also shot at 180 and 200 to compare the 70-180 and 70-200.
I then looked at the exif data from the photos. Originally I was just trying to see how different lenses compared in accuracy, especially at shared focal lengths. What I found was that I couldn't even really do this because the camera only seemed to record a couple of "fixed" values for each lens.
I don't remember the real numbers, though I could find the old thread where I losted them all, but it was like this: say that my actual distances were 20m, 40m, and 50m. I found that each lens seemed to have a short distance and a long distance that was recorded. For example, photos taken with the 70-200 at 20m and 40m would both have said 28.5 meters (regardless of what focal length I used) while the 50m shot or any shot taken from other distances over 40m all said 45.6m or something. Then on the 85mm lens it might be that anything under 24m said 13.6m while anything over said 30m. Maybe on the 180-600 it would report 37m no matter how far I really was from the subject up to 35.5m, after which it would always say 56.3m.
Again, the numbers are made up in lieu of digging up the actual numbers, but the pattern was like this. Every lens I own would only list two different distances - one for "short" and the other for "long," and then actually they all would say something like 23,000m if the subject was at or beyond the lens' infinity focus distance.
hef19898
Well-known member
The cross type AF in the R1 is somewhat overlooked it seems.
I tend to agree with the OP, and I'd like to offer my support in pointing out that it's not an uncommon complaint - you just don't see it as often as you used to because people who bring it up tend to get shouted down (not really here, where the user base is probably the best I've seen on the internet, but certainly on other forums).
I still find it works well much of the time, but there are definitely problem areas and times it just seems ro have real trouble, even with people when the light gets into "use a speedlight" territory. For instance, I don't think in 3 years I can remember any Z camera I've used - even the Z8 with the new firmwares - trying to focus on the eye or head of a flying bird. It sticks to the body pretty well, but then it's almost like the group AF from DSLR days. The value of the subject recognition for stuff like BiF is supposed to be in large part that you can focus on the eye rather than having most of the shots focus on the wing and so the head is a bit soft. I haven't had it really do this for me.
I still find it works well much of the time, but there are definitely problem areas and times it just seems ro have real trouble, even with people when the light gets into "use a speedlight" territory. For instance, I don't think in 3 years I can remember any Z camera I've used - even the Z8 with the new firmwares - trying to focus on the eye or head of a flying bird. It sticks to the body pretty well, but then it's almost like the group AF from DSLR days. The value of the subject recognition for stuff like BiF is supposed to be in large part that you can focus on the eye rather than having most of the shots focus on the wing and so the head is a bit soft. I haven't had it really do this for me.
EricKlees
Well-known member
Can't agree with the "rudimentary" subject detection. I have had little issue with subject detection. I regularly get shots friends with A1's and both the R5/R3 miss standing next to me. I have also used these other bodies and never got the same consistency as with my Z9+500PF and not the Z9+800PF
I'm not saying that it's perfect, not a 90-95% hit rate with in focus shots on a very regular basis is about as good as I can honestly expect.
The one issue that could be improved is long-necked washing birds like herons where when the neck is straight it wants to grab the body. Not if the neck is recoiled, it locks up the eye just fine. There is a pretty easy way around this but we shouldn't have to go anytime my special here
I'm not saying that it's perfect, not a 90-95% hit rate with in focus shots on a very regular basis is about as good as I can honestly expect.
The one issue that could be improved is long-necked washing birds like herons where when the neck is straight it wants to grab the body. Not if the neck is recoiled, it locks up the eye just fine. There is a pretty easy way around this but we shouldn't have to go anytime my special here
ajrmd
Well-known member
Glad to hear and would like to know more about the shooting conditions, lighting, lenses, and AF settings. My Z8 will not detect a sparrow's eye at 40 yards through vegetation short of shooting in single point and that doesn't SD. Yes, it is understandable that a camera would have difficulty distinguishing between round items and eyes. I frequently encounter this with all brands when shooting mammals particularly ungulates, primates, hippos, and elephants. The camera will frequently latch on to the ear rather than the eye. Usually, these creatures are slow enough that one has time to switch to a different AF mode.
Anyhow, back to the subject of birds, one would hope that Nikon could tweak their subject recognition to improve its results? Although a cone flower is semi-elliptical, though though the rest of the subject doesn't match. FWIW, the R5 has no difficulty detecting the bird in this scenario.
Regarding the waxwing, I was at 22.69m, in decent light against a blue sky. Yes, if the camera focused on the belly rather than the eye, it is likely that the DOF would be sufficient to render the eye in focus, but not necessarily critical focus. That being said, I was a bit surprised and again, it speaks to the opportunity to improve subject recognition.
John Navitsky
Well-known member
if you can keep a single point exactly where you need to focus, then you not only don't need subject detection, but you are probably better off without it. or rather, you are the subject detection, so use whichever works best.
as someone who shoots fast action almost exclusively, i can say that even when subject detection isn't perfect, when i do the math, subject detection is still very much a net win for me.
1) perhaps other people are much more talented than i, but with the dslr, i spent a large amount of computational overhead trying to figure where the point needed to be and keeping the subject under that dot. i'd much rather spend those mental cycles on other things.
2) related to #1, the only way you make this work is prediction. where the point needs to be on what you expect to happen. this means that when things don't happen as you expect, you simply miss. the camera can re-adjust much faster and almost always get some of a sequence.
3) even when i have focus misses in a sequence, usually the keepers are better than i could previously get (you don't put the af point on an eye of a dog running 25mph and suddenly jumping into the air, but the camera sometimes can)
4) you can still use single point, or a NxN custom point with subject detection turned off and use it like a dslr. and i tell folks, when in doubt, shoot it (exactly) like a dslr. subject detection is an additional tool you have, not a replacement.
5) you are always going to need to know the strengths and weaknesses of your different modes.
i used group a lot with the dslr. having that big, coarse, fast, spotlight of af works very well for fast moving action. however... it's also a big and coarse and absolutely not the right thing in certain cases.
with the new systems, we simply have more tools in our toolboxes, and you're still going to need to pick the right tool. knowing how each tool works (or doesn't work) in different contexts is still going to be a thing. how does the subject detection work with these types of subjects or with this kind of foreground/background clutter? how does it work with this kind of animal, or whatever is still a thing. and each brand has strengths and weaknesses across subjects and contexts and likewise, each mode of each brand does as well.
as they song says, you need to know when to hold em, and know when to... use your programmed button for single point
ajrmd
Well-known member
Eric, I appreciate your perspective and perhaps your account speaks to your skills. It would be interesting to hear more about the scenarios where you're shooting side-by-side with your friends who are using A1's, R5's, and R3's are "missing". What types of subjects and scenarios are you shooting and where are these cameras (or users) falling short? For example, I find the Z8's airplane detect to be deadly. It's cockpit recognition is the best of any body that I've shot with. Even in hazy or less than perfect air, I am astounded how it outperforms the R5/R3.
However for BIF, I'm not convinced that the Nikon algorithms are as good as the competition. Apart from the specific scenarios I illustrated (some with photos), the other area I've experienced difficulty is with osprey/eagles diving through the catch. For some reason, the AF wanders off the subject (not onto splashes or other distractions) and then comes back a few frames later. This occurs with Auto Area, Wide small, Wide Large, Wide custom, etc. If you're achieving different results in the situations I've laid out, I would appreciate learning the how's and why's.
John Navitsky
Well-known member
there is totally a lot of opportunity across all brands to improve subject detection, and i think we'll continue to see that
we see sony rolling out it's next gen af with the a9iii.
and canon's stuff with being context aware for sports.
i think we'll still see this a hot area for competition for the next few years
groob
Active member
I use AA with bird detection 99% of the time (because I’m usually shooting birds). If the AF system needs a little help, I’ll use single point to get it on the subject and then go back to AA. If I can’t get AA to lock on consistently for some reason, I’ll use 3D or just continue to use single point.
My wildlife lens is the 600mm f/4 FL E. I’d imagine there could be some differences between it and the 500mm PF, given the increased focal length and wider aperture.
My wildlife lens is the 600mm f/4 FL E. I’d imagine there could be some differences between it and the 500mm PF, given the increased focal length and wider aperture.
groob
Active member
I have certainly had an issue with long necked wading birds. But I have had good luck using 3D lately.
EricKlees
Well-known member
So for like the environment that I'm talking about one is Conowingo Dam the other is Bombay hook which is a marsh area. I consistently get better subject detection at a further distance against tree backgrounds against brown or green reeds that they fly in front of. I also get swallows at Bombay hook in front of the reeds and busy backgrounds at about 35 to 40 yards out without too much issue in Bombay hook there's lots of ducks I have no problem with latching on to the eye whereas friends with the A1 there's 10 to grab the eye of the reflection not the actual bird.
Bird and flight if I'm having a little bit of an issue because it's at distance going in the DX mode latches right onto the head or eye with no issues and if I'm that far out I'm going to be cropping anyway so I'd rather have a text sharp image at 19.2 megapixels rather than potential missed focus on a 45 megapixel image.
There are also some techniques that you can do to easily help the subject detection. So my skill is one thing learning your equipment is another but like I said I've used all of those cameras and found that I consistently hit focus at a higher rate and more consistency with my Nikon Z9..
There's also little doubt that I do get better results than others users much of the time I have about 10 or 12 friends that have a z9. But it goes towards learning how to get the most out of your equipment. One of those is knowing how to leverage The passive AF to your advantage. This is one of the things that I teach in my z9 techniques for people
John Navitsky
Well-known member
EricKlees
Well-known member
I have my rsf set to change my subject detection from bird to animal and my default AF mode from Auto area back to the wide c1 and I'll just hand off to either 3D or even Auto but that gets me around that and then when I'm done with that I'll just go back to my normal bird subject detection with auto area AF mode
EricKlees
Well-known member
I call when you have a few random missed focus frames in a burst sequence, AF drift.if you can keep a single point exactly where you need to focus, then you not only don't need subject detection, but you are probably better off without it. or rather, you are the subject detection, so use whichever works best.
as someone who shoots fast action almost exclusively, i can say that even when subject detection isn't perfect, when i do the math, subject detection is still very much a net win for me.
1) perhaps other people are much more talented than i, but with the dslr, i spent a large amount of computational overhead trying to figure where the point needed to be and keeping the subject under that dot. i'd much rather spend those mental cycles on other things.
2) related to #1, the only way you make this work is prediction. where the point needs to be on what you expect to happen. this means that when things don't happen as you expect, you simply miss. the camera can re-adjust much faster and almost always get some of a sequence.
3) even when i have focus misses in a sequence, usually the keepers are better than i could previously get (you don't put the af point on an eye of a dog running 25mph and suddenly jumping into the air, but the camera sometimes can)
4) you can still use single point, or a NxN custom point with subject detection turned off and use it like a dslr. and i tell folks, when in doubt, shoot it (exactly) like a dslr. subject detection is an additional tool you have, not a replacement.
5) you are always going to need to know the strengths and weaknesses of your different modes.
i used group a lot with the dslr. having that big, coarse, fast, spotlight of af works very well for fast moving action. however... it's also a big and coarse and absolutely not the right thing in certain cases.
with the new systems, we simply have more tools in our toolboxes, and you're still going to need to pick the right tool. knowing how each tool works (or doesn't work) in different contexts is still going to be a thing. how does the subject detection work with these types of subjects or with this kind of foreground/background clutter? how does it work with this kind of animal, or whatever is still a thing. and each brand has strengths and weaknesses across subjects and contexts and likewise, each mode of each brand does as well.
as they song says, you need to know when to hold em, and know when to... use your programmed button for single point
I found that a couple settings/features turned off help with this. 1st, shoot in a flat picture control. If you have the custom picture control that gives the blown highlight "blinkies", this is why it's based on flat.
The other thing that I've found to help with less AF drift is turning off the shooting indicators (called release timing indicators) in the EVF that flash to let you know visually that you're actively taking photos. Not sure why but could be as simple as less fo r the EXPEED 7 to deal with.
Last edited:
It hasn't shipped yet so we will see if it lives up to the hype.
John Navitsky
Well-known member
John Navitsky
Well-known member
it’s been unclear to me if cross type sensors are problematic for mirrorless sensors or if it’s just the way things have evolved to this point.
i wonder if it’s because canon did more of the sensor development is why they did cross types (and maybe why it’s taken as long as it has). conversely i wonder if the iterative development of the majority of sensors has made it harder to include cross types (maybe they’d have to redesign too much to be practical).
Last edited:
John Navitsky
Well-known member
we should also remember that (afaik) cross type sensors will improve af, not subject detection
Hmm. I do not get subject distance with any of my Z8 or Z9 files. Even when I process through NX Studio instead of my preferred Capture One Pro.
John Navitsky
Well-known member
so you’re running with zero visual indicators?
ajrmd
Well-known member
Like you, I spend a reasonable amount of time (in season) at both Conowingo and Bombay Hook and am surprised we never met. Perhaps, we'll be fortunate to connect up this fall? So, what I am particularly frustrated with is what I think you describe as "AF drift". Initial capture is achieved and then when the eagle strikes, the AF seems to drift off target for a few frames, only to reacquire again after the strike. These are head scratching sequences where one is panning a strike usually less than 50 yards away and the target is large in the frame (nearly occupying 80+% of the vertical field. How or why this occurs with my Z8 is simply perplexing as again, I've not experienced this with Sony/Canon. If they acquire AF, they don't seem to lose in through the strike. With respect to the reflection "latching", I've encountered this on rare occasion with the Z8 though if one "pumps" the AF, i.e. releases and reacquires, this usually resolves.
The thought of shooting a flat picture control is intriguing though it is difficult to see why that would affect AF? Conceivably, I could see how using the custom picture control which produces blinkies could tax the EXPEED 7 chip, though why would NL (what I use) rather than "flat" lead to drift?
The reason it might make a difference is that the camera uses the EVF video feed to perform AF operations including subject detection, and the image shown in the EVF video feed has whatever picture control is set applied. Try it right now if you like: set picture control to flat, point the camera at something, and look through the EVF or on the rear LCD: you'll see that the image is, well, flat, with washed out colors, lower texture, etc. Now set it to vivid and look again to see everything bright and colorful. Etc.
The idea behind setting the picture control to help with AF is that by increasing contrast in the image shown in the EVF display, there will be more contrast for the AF to use for AF. Contrast along edges can be increased by increasing the actual contrast slider or the sharpening or both.
I've mentioned this to him before, but Eric's suggestion to use flat is odd to me, because it is as far as I can tell directly contradictory to the pretty widely accepted advice that went around regarding the Z9 closer to its launch and which has been suggested by a number of "expert" personalities on Youtube, etc. The reason for this is that the flat picture control actually lowers contrast and sharpening compared to other picture controls and so following the theory of it all it should decrease AF performance, not help it. In fact, the common advice that you will find looking through old threads is to manually edit the picture control the push sharpening and contrast up, while the flat picture control decreases them. I did after the last time Eric mentioned this go ahead and try switching to the flat picture control just to see what happened and I found that I had a noticeably more difficult time with the AF.
EricKlees
Well-known member
Correct. I rarely shoot silent (I have FN2 set to toggle silent) and have the sound set to level 2 Type B and only use it if I am super close to a subject
EricKlees
Well-known member
Might have bumped into each other and just never introduced ourselves. Anyway we'll have to try and hook up at some point.
With the AF box if it comes off the subject while you're painting and shooting I continue to shoot and I tell people to continue to shoot as if it's still on. Nine times out of 10 you're still focusing on your subject but the AF Point can lie behind a bit. They've improve this a bit with I want to say firmware 3.0 maybe I can't remember which but it can still lag a bit but the camera is still locked on and continuing the track the subject.
As far as Eagle coming down to fish and the z9 losing focus at the point of impact I've had a couple of friends of mine say they've had that issue but most don't and I usually don't I mean I can show you image after image after image where before during and after the pickup is tax short doesn't get distracted by the splash or anything of that nature. I even get ospreys as they dive into the water when they're under and coming up and have like that water helmet so to speak it's still locked onto the eye.
But I get osprey's fishing Eagles fishing routinely within 50 yards so not much of an issue and I've had them as far as 250 yd.
The flat picture control helps the autofocus in general. I find that I get less AF drift by turning off the release timing indicator which is D15 in the custom menu or the one with the pencil. The flat picture control I honestly don't remember but I read that sometime in 2022 and the reasoning made sense but I can't remember what it is at this point but I do see a difference compared to using standard or neutral. The custom picture control doesn't cause really any issues there as it's based on flat. And in post I tend to like to use either neutral or the landscape camera matching profile better for the whites especially for the eagle heads
- Status