Another rant about Fujifilm af-c tracking with eye/face detect

He do not say anything about the AF-C Settings, the setting for energy level and the settings for AF+MF, which are relevant for a functioning focuspriority.

My X-H2s with Firmware 6.0 and the right settings does not show the behavior, he shows in his video. Works fine without such issues.
Otherwise we all know, that the menue of the Fujifilm cameras has many traps and silly defaults. So I think he and other people, who say they have such issues, are lost in space of Fujifilm menues.

That is the biggest issue in my opinion and that is to blame in public, that defaults are crooked and there are hidden interactions between functions like AF+MF and focuspriority.
He show his AF-C custom settings here:
So, You are saying that changing in camera menu: power management -> performance: boost to normal (or economy) will cause, that camera AF system after choosing face/eye detection will instantly loose tracking whenever any object will show in the frame?
Same behavior for changing AF-C priority selection from focus to shutter? :-O
Yes, I think he has such issues with boost or AF+MF ON, which cancels focuspriority. I have no other idea. I have testet it again myself with my wife in the last hour. Eye AF green and focus on eye and not on the hand. Perhaps his camera is defect?
 
It's quite simple. As long as there is one or more users for whom the camera works without problems in the scenario shown, it is crystal clear that it can never be a problem with the camera. Regardless of whether 5 or 10 others have a similar problem. It simply proves that it works. And that those who have the problem have not configured something perfectly in the settings.
 
It's quite simple. As long as there is one or more users for whom the camera works without problems in the scenario shown, it is crystal clear that it can never be a problem with the camera. Regardless of whether 5 or 10 others have a similar problem. It simply proves that it works. And that those who have the problem have not configured something perfectly in the settings.
Fair enough, it makes sense - under condition, that it is rally true.
Like i said before, I will try to do such test with my X-H2 and Viltrox 27/1.2 after the weekend (but no promises)
Then I will write down my findings.
 
He do not say anything about the AF-C Settings, the setting for energy level and the settings for AF+MF, which are relevant for a functioning focuspriority.

My X-H2s with Firmware 6.0 and the right settings does not show the behavior, he shows in his video. Works fine without such issues.
Otherwise we all know, that the menue of the Fujifilm cameras has many traps and silly defaults. So I think he and other people, who say they have such issues, are lost in space of Fujifilm menues.

That is the biggest issue in my opinion and that is to blame in public, that defaults are crooked and there are hidden interactions between functions like AF+MF and focuspriority.
He show his AF-C custom settings here:
So, You are saying that changing in camera menu: power management -> performance: boost to normal (or economy) will cause, that camera AF system after choosing face/eye detection will instantly loose tracking whenever any object will show in the frame?
Same behavior for changing AF-C priority selection from focus to shutter? :-O
Yes, I think he has such issues with boost or AF+MF ON, which cancels focuspriority. I have no other idea. I have testet it again myself with my wife in the last hour. Eye AF green and focus on eye and not on the hand. Perhaps his camera is defect?
Alright, You do that and please write down Your results :)
 
I've tested this myself and can confirm that I experience such issues with the X-H2S - the camera will focus on hands, fingers, or anything in front of the face, even though the green box on the LCD screen consistently sticks to the eye.

However, I think the deciding factor here is the "Release Priority" versus "Focus Priority" setting. Changing from Release Priority to Focus Priority eliminates this issue immediately.

This does not mean that the firmware isn’t flawed. I believe it is. The X-H2S shouldn’t focus on a hand while displaying a sticky green box around the subject's eye.
 
I've tested this myself and can confirm that I experience such issues with the X-H2S - the camera will focus on hands, fingers, or anything in front of the face, even though the green box on the LCD screen consistently sticks to the eye.

However, I think the deciding factor here is the "Release Priority" versus "Focus Priority" setting. Changing from Release Priority to Focus Priority eliminates this issue immediately.

This does not mean that the firmware isn’t flawed. I believe it is. The X-H2S shouldn’t focus on a hand while displaying a sticky green box around the subject's eye.
That is interesting.
So we have first clue, why this is happening.
Thank You for Your report!
Can someone with X-H2s and newest Fuji primes test it with release priority vs shutter priority, and if changing this setting fix false AF confirmation issue?
 
I'm always a little puzzled about this type of video.

Face / eye detection was absolutely horrible on my X-T2 to the point that it saw faces on the walls, trees, clouds... but absolutely not on people's faces.

What did I do? Well I took advantage of the wide/tracking AF all in AF-C and never had any issues again with the X-T2's autofocus.

When it comes to newer Fujis, I don't have much experience with them, but I handled some in a store : the tracking autofocus is brilliant and works very well, about as well as my current Nikon Z camera. I don't think this would be an issue for anyone to use.

I can understand the critics : yes the AF is not as good as Sony, Canon, Nikon, Lumix... but at teh same time it's not that far and there is always workarounds. Nowadays you almost ahve the impression that people just set up the camera for automatic subject detection and then lash out on the camera when it does not have a perfect hit rate.

If anything using older cameras has taught me anything is that not leaving the camera do all the work for you is a healthy recommendation that will make you react to problems better. Otherwise the moment your subject detection is buggy and doesn't work you're SOL.
 
I've tested this myself and can confirm that I experience such issues with the X-H2S - the camera will focus on hands, fingers, or anything in front of the face, even though the green box on the LCD screen consistently sticks to the eye.

However, I think the deciding factor here is the "Release Priority" versus "Focus Priority" setting. Changing from Release Priority to Focus Priority eliminates this issue immediately.

This does not mean that the firmware isn’t flawed. I believe it is. The X-H2S shouldn’t focus on a hand while displaying a sticky green box around the subject's eye.
That is interesting.
So we have first clue, why this is happening.
Thank You for Your report!
Can someone with X-H2s and newest Fuji primes test it with release priority vs shutter priority, and if changing this setting fix false AF confirmation issue?
Personally, I'd like to see a video in which the H2S is used in DR100, AF-C eye detect, AF-C custom set 2 (ignore obstacles) at default settings, to photograph a subject in good light (ISO no higher than 200) with focus lock acquired before the subject brings their hand into the outer part of the frame and slowly moves the hand into a position where it might confuse focus acquisition.

If the person making the video wants to try other AF-C custom sets, experiment with release modes, try different AF-C modes (zone, single point), and different models those would be plusses. But the first order of business is to document AF-C eye detect behavior under optimal conditions using default settings.

A question I have about the subject and eye detect settings is, are they more or less effective depending on the subject's appearance. There's just one human face & eye detect mode. But we live in a beautifully diverse world. We don't all look alike. Is eye detect equally effective & reliable regardless of the model's skin tone and facial structure?
 
Personally, I'd like to see a video in which the H2S is used in DR100, AF-C eye detect, AF-C custom set 2 (ignore obstacles) at default settings, to photograph a subject in good light (ISO no higher than 200) with focus lock acquired before the subject brings their hand into the outer part of the frame and slowly moves the hand into a position where it might confuse focus acquisition.
These are two still images - animal subject detection rather than human face/eye detection but easier to see the effect because of the long dog node.

The images were taken one after the other in focus priority mode with the camera (and EXIF) confirming eye detection had been acquired.



2ba6835d911942cc90b2152cdfebbda2.jpg



8cc15a233030417894988708b88c31e9.jpg
 
Personally, I'd like to see a video in which the H2S is used in DR100, AF-C eye detect, AF-C custom set 2 (ignore obstacles) at default settings, to photograph a subject in good light (ISO no higher than 200) with focus lock acquired before the subject brings their hand into the outer part of the frame and slowly moves the hand into a position where it might confuse focus acquisition.
These are two still images - animal subject detection rather than human face/eye detection but easier to see the effect because of the long dog node.

The images were taken one after the other in focus priority mode with the camera (and EXIF) confirming eye detection had been acquired.

2ba6835d911942cc90b2152cdfebbda2.jpg

8cc15a233030417894988708b88c31e9.jpg
To see what really happened:

Either please provide the raws (not jpegs) or you can do it yourself by going to https://www.solentsystems.com/exif3/ then loading each image.

This will show you where the focus was detected, then you can compare it with the focus you see in the image. Also check on the right panel for the AF section with all the AF settings used.

--
Thanks
The Dogs of Vancouver, BC
 
Personally, I'd like to see a video in which the H2S is used in DR100, AF-C eye detect, AF-C custom set 2 (ignore obstacles) at default settings, to photograph a subject in good light (ISO no higher than 200) with focus lock acquired before the subject brings their hand into the outer part of the frame and slowly moves the hand into a position where it might confuse focus acquisition.
These are two still images - animal subject detection rather than human face/eye detection but easier to see the effect because of the long dog node.

The images were taken one after the other in focus priority mode with the camera (and EXIF) confirming eye detection had been acquired.

2ba6835d911942cc90b2152cdfebbda2.jpg

8cc15a233030417894988708b88c31e9.jpg
To see what really happened:

Either please provide the raws (not jpegs) or you can do it yourself by going to https://www.solentsystems.com/exif3/ then loading each image.

This will show you where the focus was detected, then you can compare it with the focus you see in the image. Also check on the right panel for the AF section with all the AF settings used.
I had expected that the camera had missed the eye and instead detected the face - in which case focusing on the nose would have made sense - but that isn't the case - in both cases the EXIF reports that the eye has been found and the focus point is reported as being on the eye


 
I have found the eye AF pick up noses and ears which it perceives as eyes. Especially when the eyes are not open or straight at the camera. It does not happen often enough to bother me though.
 
I can replicte the issue on my X100VI (didn´t test on my X-T5) if i´m in Zone AF set to Front with Release Priority. Eye detect enabled, obviously. Tested just photo mode.
 
Last edited:
Personally, I'd like to see a video in which the H2S is used in DR100, AF-C eye detect, AF-C custom set 2 (ignore obstacles) at default settings, to photograph a subject in good light (ISO no higher than 200) with focus lock acquired before the subject brings their hand into the outer part of the frame and slowly moves the hand into a position where it might confuse focus acquisition.
These are two still images - animal subject detection rather than human face/eye detection but easier to see the effect because of the long dog node.

The images were taken one after the other in focus priority mode with the camera (and EXIF) confirming eye detection had been acquired.

2ba6835d911942cc90b2152cdfebbda2.jpg

8cc15a233030417894988708b88c31e9.jpg
To see what really happened:

Either please provide the raws (not jpegs) or you can do it yourself by going to https://www.solentsystems.com/exif3/ then loading each image.

This will show you where the focus was detected, then you can compare it with the focus you see in the image. Also check on the right panel for the AF section with all the AF settings used.
I had expected that the camera had missed the eye and instead detected the face - in which case focusing on the nose would have made sense - but that isn't the case - in both cases the EXIF reports that the eye has been found and the focus point is reported as being on the eye

https://www.dropbox.com/scl/fi/9zwm...ey=8owziamtfqhr3fez1thvocomj&st=1wotc5ln&dl=0

https://www.dropbox.com/scl/fi/xvso...ey=qep1kigcp2hhntc5c6mgcyeww&st=aq4b4j2f&dl=0


6c2d883213c944c592f46fdeb9ecff35.jpg

and



79c76e4bd9d64351a0e085e009d25baa.jpg

This looks like you are using the settings from Morris0 but they are not the best for this situation. His settings are "tailored" for BIF.

With resting dogs, case 1 would work better, especially with cats and dogs, especially sensitivity zone: center instead of front.



--
Thanks
The Dogs of Vancouver, BC
 


This looks like you are using the settings from Morris0 but they are not the best for this situation. His settings are "tailored" for BIF.

With resting dogs, case 1 would work better, especially with cats and dogs, especially sensitivity zone: center instead of front.
He was using Wide/Tracking. Doesn´t center/auto/front only apply if you´re in Zone AF?
 
This looks like you are using the settings from Morris0 but they are not the best for this situation. His settings are "tailored" for BIF.

With resting dogs, case 1 would work better, especially with cats and dogs, especially sensitivity zone: center instead of front.
He was using Wide/Tracking. Doesn´t center/auto/front only apply if you´re in Zone AF?
No, not my understanding. It applies to the box to the box with the subject found.
 
Personally, I'd like to see a video in which the H2S is used in DR100, AF-C eye detect, AF-C custom set 2 (ignore obstacles) at default settings, to photograph a subject in good light (ISO no higher than 200) with focus lock acquired before the subject brings their hand into the outer part of the frame and slowly moves the hand into a position where it might confuse focus acquisition.
These are two still images - animal subject detection rather than human face/eye detection but easier to see the effect because of the long dog node.

The images were taken one after the other in focus priority mode with the camera (and EXIF) confirming eye detection had been acquired.

2ba6835d911942cc90b2152cdfebbda2.jpg

8cc15a233030417894988708b88c31e9.jpg
To see what really happened:

Either please provide the raws (not jpegs) or you can do it yourself by going to https://www.solentsystems.com/exif3/ then loading each image.

This will show you where the focus was detected, then you can compare it with the focus you see in the image. Also check on the right panel for the AF section with all the AF settings used.
I had expected that the camera had missed the eye and instead detected the face - in which case focusing on the nose would have made sense - but that isn't the case - in both cases the EXIF reports that the eye has been found and the focus point is reported as being on the eye

https://www.dropbox.com/scl/fi/9zwm...ey=8owziamtfqhr3fez1thvocomj&st=1wotc5ln&dl=0

https://www.dropbox.com/scl/fi/xvso...ey=qep1kigcp2hhntc5c6mgcyeww&st=aq4b4j2f&dl=0
6c2d883213c944c592f46fdeb9ecff35.jpg

and

79c76e4bd9d64351a0e085e009d25baa.jpg

This looks like you are using the settings from Morris0 but they are not the best for this situation. His settings are "tailored" for BIF.

With resting dogs, case 1 would work better, especially with cats and dogs, especially sensitivity zone: center instead of front.
I’m not using anyone’s settings - but that isn’t the point - the camera has found the eye - but at random chooses to focus on the head
 
This looks like you are using the settings from Morris0 but they are not the best for this situation. His settings are "tailored" for BIF.

With resting dogs, case 1 would work better, especially with cats and dogs, especially sensitivity zone: center instead of front.
He was using Wide/Tracking. Doesn´t center/auto/front only apply if you´re in Zone AF?
No, not my understanding. It applies to the box to the box with the subject found.
When i had a X-Pro2, X-T2 and X-T3, i did some tests to see if center/auto/front settings applied to single, zone, and wide. At that time i found that they only applied to zone. But then i didn´t use any face/eye detection with those cameras (found it lacking).

As i mentioned on a previous post, i´ve tested the X100VI, and i could replicate the issue being discussed while using eye detect, if i set the camera to zone with front (not center) and release priority. If i set the camera to front but in single point af or wide/tracking, even with release priority, i couldn´t replicate the issue.
 
Last edited:

Keyboard shortcuts

Back
Top