Who still thinks face detection AF is a gimmick after having used it for a while?

[New Daddy]

(Clarification: "face detection AF" means the ability of the camera to lock focus on a person's face. I don't mean the camera's ability to recognize the facial feature and tie it to a specific person stored in the camera's memory. Although the two are related, I'm solely discussing "face detection AF" here.)

At first, I didn't think much of face detection AF. But it turns out that it is really convenient and time-saving. It basically takes care of a lot of the recomposition and metering issues when you are shooting a person. You don't have to center-focus-lock-and-recompose or move the AF point to an off-center face of your subject. When the camera has face detection AF, it will bias for the face in metering too. With face detection AF, my long routine of spot-meter-off-face-and-AE lock, followed by lock-AF point-on-face-and-recompose is eliminated.

There is a caveat, of course. It doesn't always work, especially when your subject is not looking at the camera or when your subject is surrounded by other people. (Face "recognition" comes in really handy in the latter type of situation.)

So any veterans who have found this new-fangled technology convenient too?

[rpcrowe]

I prefer to select my own focus points. The 7D has a wonderfully capable auto-focus system which I can modify to just about any way I desire...

[PhotomanJohn]

Sorry Rick, I'm with Richard. I pick the focus point and use evaluative metering on average scenes. That being said, if it works for you then that is a good way to go. You will continue to get it better figured out and know when you can't trust it.

John

[Stagecoach]

Rick,

I have never used it but seem to recall my camera has something like this. How do you find it works at the different apertures and nailing focus on the eyes?

Grahame

[Shadowman]

I like how it will detect a smile from another photograph or poster.

[New Daddy]

Rick,

I have never used it but seem to recall my camera has something like this. How do you find it works at the different apertures and nailing focus on the eyes?

Grahame

Mine doesn't focus on the eyes, just on the face. I think the latest Olympus can nail focus on the eyes, or an eye as the user can choose either the left eye or right eye for priority.

I don't use razor-thin DOF for environmental portrait, so the face detection has worked pretty well.

[New Daddy]

I'd like also like to emphasize that I'm talking about casual, run-and-gun type of shooting. If this was a studio setting, I too would choose my own focus point, etc.

The metering bias for focus point in face detection is really convenient and quick too. Relying on the rule of thumb for adjusting the exposure based on the skin tone of the subject (+07-1 EV for caucasian, 0 for asian and -1 EV for african-americans), metering becomes really quick with face detection. I've checked all my cameras, and every camera with the face detection had metering bias for the face. So, "face detection" is not just an AF technology, but it's rather a whole package - AF and metering for faces.

Again, I wouldn't say face detection trumps all other methods, but in run-and-gun situation, I'll bet I'll have far better chance of capturing a critical moment with face detection.

[GrahamH]

I like the way that the question is posed, I find it entertaining.

You have to have used face detection for a while, then assess it's degree of gimmickness.

Apart from reviewers, who is going to use a feature that they DON'T like for a while?
The question itself tends to be self serving to provide a positive result.
Those who don't like it, or use it aren't really in a position to assess it's gimmickness.
Me, never used it. Never found a case where I had cause to do so. As such I cannot assess if it is a gimmick or not.
In line with most other postings, I use more traditional techniques.
It's harder than people think to ask a suitable question like this.
Graham

[Inkanyezi]

There are several caveats to face detection, and to understand the underlying technical aspects, it is worth the while to assess the technical properties of the function and what possibilities it offers for the image.

There is a vast difference between DSLR and other types of digital cameras, when it comes to face detection. It could be described as simple as a DSLR cannot detect a face and set proper focus, while other cameras can. There's where opinions may arise, and of course it is a bit more complicated, but in broad lines, the DSLR cannot (to a large degree), because in its viewfinder, mostly there are no sensor elements with enough spatial density to identify a face.

Nikon was the first DSLR manufacturer to include a small digital camera within the viewfinder, whose primary purpose was to distinguish between various fields in the view, to integrate light measurement. This tiny camera also has the ability to identify faces, in order to measure light for them, and possibly to focus on the face. Those little help cameras within the viewfinder come with different densities, one with fewer fields for measurement in the cheaper models and in the more expensive models, it has larger density and therefore is more capable.

But then to focus, other sensors are used, placed below a little relay mirror behind the main mirror that has a semi-transparent spot before that relay mirror. At first DSLR cameras mostly had just one of those spots, the centre one, "cross-wise", so that it would focus on horisontal, as well as vertical lines. For focusing on an eye, also the linear ones are good enough, but to focus at all, the target must hit the sensor. So in the older, and in the cheaper cameras, those spots are few and widely spaced, and do not extend far from the centre of the sensor toward the sides and corners.

So in order for face detection to function at all, the camera must:

1. identify the face with the little VF camera
2. have the crucial feature of the face (the eye) within the AF sensor coverage

The first point requires that the face is large enough in the image to be identified as a face by the low resolution camera in the viewfinder. The second point requires that the crucial feature, the eye, can be identified as well, and that it will hit one of the AF sensors.

So there's where EVIL and other mirror-free cameras have an edge over DSLR. They can identify the face, because their density of measuring elements is much higher. They can also identify the crucial feature and focus for it, as those cameras do not have focusing spots but can use the entire sensor area. The drawback is that with many more elements to evaluate, it uses more computer power and takes more time. Therefore, cheap compact cameras don't care for focusing on the eye, and they use fewer of the light wells on the sensor for their assessment, while some cameras with more computer power use more of the sensels and can do it with more precision. There are cameras that not only identify the face, but also focus on the eye, also when those features do not cover a large part of the viewfinder area. None of those is a DSLR.

So those are the reasons why these "gimmicks" are mostly not present in DSLR cameras, but in compacts and also come in the mirror-free systems.

[New Daddy]

I like the way that the question is posed, I find it entertaining.

You have to have used face detection for a while, then assess it's degree of gimmickness.

Apart from reviewers, who is going to use a feature that they DON'T like for a while?
The question itself tends to be self serving to provide a positive result.
Those who don't like it, or use it aren't really in a position to assess it's gimmickness.
Me, never used it. Never found a case where I had cause to do so. As such I cannot assess if it is a gimmick or not.
In line with most other postings, I use more traditional techniques.
It's harder than people think to ask a suitable question like this.
Graham

When you take a moment and think about it, you'll realize that your statement applies to just about every question, unless the question welcomes rash conclusions.

[GrahamH]

When you take a moment and think about it, you'll realize that your statement applies to just about every question, unless the question welcomes rash conclusions.

Hi Rick (or should that be Hey - sorry old joke).
Actually, it was the inclusion of the qualifier that makes the question self-serving. If the qualifier did not create a bias, then the question would be less likely to be self-serving. The respondee would then have the option to state their degree of use of the 'feature/gimmick'. There are many studies and questionnaires that are flawed as a result of the way the questions are asked that produce a basis.
Thus, 'Who still thinks face detection AF is a gimmick?' is considerably less likely to produce bias than the way it was stated.
Not picking to be mean, just having some fun, keeping it light. No offence intended.
Most questions I ask (that I am aware of - and I am aware that I may be unaware ) are designed to minimise or eliminate bias. However, the listener is always in the position to add-on interpretations of their own and create a totally different question to that intended (and never mind ambiguities).

Back to the question, after a quick scan, I don't see anyone who responded who HAS used the feature 'for a while' and who has responded that they consider it a gimmick. Love to hear from someone who HAS actively used it and considers it a gimmick. I would love to know WHY they continued to use it, even after they formed their opinion.
I would suggest that the occasional use of a feature when considered necessary means that the feature is not a 'gimmick' merely a infrequently used feature. For me, I rarely used BULB, but I don't consider it a gimmick.
I haven't used face detection sufficiently to be able to form an opinion and so haven't put it into the gimmick designation.

Graham
bad day

[ajohnw]

I assume this is purely in the realms of contrast AF on the full sensor plus electronic viewfinders. I tend to leave mine on but do see it as a bit of a gimmick. For one it can recognise several faces but not as many as may be in front of the camera so in that sense it doesn't really help I still need to come to some conclusion as to where to focus. 2ndly I can put a focusing square where ever I like. Having the face detection do that automatically speeds that up a little but in real terms I may as well leave it some where and position the square and 1/2 press the release etc. 3rdly it doesn't always work as it can have problems with backgrounds where as a normal focusing square generally works. 4thly I have little interest in exposing purely on the basis of some ones skin tones. Especially on this one as by nephew is 50% Spanish. I also may want a shot like this that retains whites and blacks. A guessed face exposure would be too far out.. Not my best PP just a couple of mins but all that needed lifting was the blacks a little.



Edit now I have considered what it has done for me it's going to get turned off.

John
-

[Inkanyezi]

I leave it turned on. When I use µ4/3 lenses it works, and it focuses at the eye (OM-D). The Panasonic G1 did not focus on the eye, but rather at the hair by the temples, somewhat before the ears. On that camera I had to turn it off. I find it a bit amusing that the function shows face detection also when legacy MF lenses are mounted, and sometimes it finds faces where there aren't any. However I have left it on so far, and it hasn't ruined anything. Whenever I set a focus spot, the camera focuses just there, even when a face may be identified somewhere else. Those green face detection squares still turn on, but the camera focuses on my selected spot. I think the feature is handy when taking pictures of people.

[RobertsMx]

Face Recognition can be a useful tool in conjunction with Face Registration. Here is a video demonstration of object tracking with Sony A77 by Gary Fong.

Sony has just added Eye AF as well (to A7, A7r and RX10) and I believe object tracking is expanded to eye (not just face).