Where do I even begin? *big post*

[bitsnpieces]

No really, I don't even know if this is the correct section to put this in, so if not, feel free to move to proper section if required - sorry for the inconvenience.
I think it belongs here just because I feel 'new' and want to know more about this/need help with this.

Anyways, this is a bit of questions, rant, comments, etc, thus, where do I even begin? And there's a bit to read.

I guess, straight to the point, how do I get my Sony a65 to have better auto-focus?
Now, that's a very vague question, and here's why (thus the comment/rant part):

I just finished photographing a sports night at church today - they played basketball and volleyball (I'll be going through the photos as time permits to crop, enhance, etc, which will take a while), and I was using continuous auto focus with wide AF area, but I found some oddities.

1) The camera seems to look at everything as a flat scene, and likes to focus in the centre (I've also tried zone which helped marginally). Now, I've read somewhere where someone noticed that different AF points in the AF area focused quicker than other points, nevertheless, my camera really likes the centre (such as when I'm taking a group photo, it focuses on the centre AF point, or around there - there isn't even multiple points focused, just one AF point). But, when there's movement, the camera 'may' pick it up, and focus on one of the points other than the centre. Even so, most of the time, it still focused on the centre, rather than the 'people running around'.

So I guess, what can I do to make the focusing more intelligent... I guess, if that's the right word to use... To track people more (other than using the tracking function because then it's only focused on one thing, and without having to pick the points manually) so that I can get a broader focus?
Could it be my aperture is too wide, so it doesn't have the breathing room to focus on multiple points? I was using f5 throughout the shots, hoping there would be enough to focus with, but not too small so that I can still use 1/160 and ISO 3200.

2) I'm still trying to figure out the face detection feature. For example, on point and shoots, it detects the face, it'll focus on the faces, snap, walah - done. I've noticed on both my Sony a65 and a55, it doesn't do that. It finds the face, a box encompasses them, but it'll still focus on one of the 15 AF points. So although the head and box is just above one of the points, it focuses on the point instead, where the neck/collar/shoulder is (such as the problem previously mentioned when I'm taking a group picture).
How do I get it so that when it detects a face, it focuses to that face? I've looked in my settings and don't see anything which could be preventing it from doing so, other than not using smile shutter to automatically take shots when it detects a smile (I do have face detection on nonetheless).

One thing I noticed though is, if the face/box happens to be on one of the AF points, sometimes it'll focus on that head. Sometimes it'll just focus on the AF point (if the box is only partially in the AF point) - so I don't know if getting the box to focus requires it being on an AF point or not... which I find defeats the purpose of face detection...
Do I need to be in a different mode like auto or portrait? I was using manual for the night, before going to aperture priority to let the camera try and pick it's own shutter speeds to see how it'd go.

So those are pretty much my concerns.

Other than that, one reason for tonight was to test the batteries, auto focus, and speed I guess?

We played indoors, just under room temperature or so around there, GPS off, quick review off, used LCD rather than viewfinder, no flash, ISO 3200, f5, 1/160 (average), JPEG (fine), generic battery, image stabilisation on, and continuous auto focus.
I managed to take 1700 photos of consistent burst shots, only chimped a handful of times, and still have 51% battery left according to the camera (started around 96% or something).

The camera specs show it only takes around 500 photos per charge, so with this difference, I'm wanting to know a little more about the process of how they test their batteries?
I want to know so I can gauge whether having 2 batteries will be enough for when I shoot someone's engagement photos, wedding, etc, or not.
Hopefully then I can get a better idea of how much power different functions use also.

I generally don't use GPS, I usually do use the viewfinder more often then I did tonight, and I generally don't burst. During casual strolling for taking photos, I chimp a lot to save having to do that after I get home (I know, you should always be ready to shoot - it's a very casual stroll).

Anyways, any comments/help appreciated. Ta. :)

[arthurking83]

.....

I guess, straight to the point, how do I get my Sony a65 to have better auto-focus?
Now, that's a very vague question, and here's why (thus the comment/rant part):

I just finished photographing a sports night at church today - they played basketball and volleyball (I'll be going through the photos as time permits to crop, enhance, etc, which will take a while), and I was using continuous auto focus with wide AF area, but I found some oddities.

..
ship
..


Other than that, one reason for tonight was to test the batteries, auto focus, and speed I guess?

We played indoors, just under room temperature or so around there, GPS off, quick review off, used LCD rather than viewfinder, no flash, ISO 3200, f5, 1/160 (average), JPEG (fine), generic battery, image stabilisation on, and continuous auto focus.
I managed to take 1700 photos of consistent burst shots, only chimped a handful of times, and still have 51% battery left according to the camera (started around 96% or something).

....


OK, until some uber knowledgeable sports person comes along .. you have me to bore you to tears!

In the quoted text of yours, you already have two points that should be pointed out, so that you can understand what has happened and what to look for next time.

Firstly, you can't get the a65 to have better AF than it already does. Well I don't know of any, and have no knowledge of Sony firmware hacking that could help .. and I doubt that any exists.(others can chime in with real answers tho!)
So, your question should have been more like "how can I get maximum efficiency or success rates from the a65's AF system!"

If that's not totally understood.... basically, the a65 will only focus as good as it does.


How can you get it t work better than it did for you on the day?
Assuming that it was a bit of a problem for you, the basic concept that should be taken from (at least) what I write here, is that the photographer gets the photos .. the camera is basically a hammer that has the ability to capture pictures nicely(when the photographer does a half decent job). You can of course bring the wrong tools to the job too tho. You don't want a lightweight dressmakers 100g platinum hammer to drive a 5" nail into WA hardwood .. do 'ya!

So as long as you have the notion that the photographer does the work, I'll try to point out why I think the camera was used wrongly too now.

First you said you tried many various AF features! You shouldn't. You should use a single AF point, you should stick to that AF point and have a probability already in mind that the image may need to be cropped to make that image look .... <in a certain way> The AF point could be any, most like to use centre, as I do too sometimes, but I also like to use off centre AF points a lot too.
Using off centre AF points can get good results, and any good images need zero to very little cropping, whereas some images, to look interesting shot with the centre focus point may need lots of cropping to give good proportions(in the image).
This all depends on taste and style .. no hard a fast rule tho ... the rule you should try to stick too is single af point(on the a65). You follow the action and keep that af point pointed at the main attraction.

You need to remember at the market level that the a65 is aimed towards. It's more like families, and young people on holidays around the world and gooey stuff like that.
it doesn't have a very 'pro like' af system. You shouldn't expect too much from it. It may even excel at what you tried at the sports night compared to other cameras in it's market segment, but it's not a 1dMkV or D4s(both high end pro cameras!!)
So can you improve the a65's focusing system .. actually you could, by getting an a77. Ok this is cheating, but, hopefully, you get the point.

cheaper end consumer camera in possibly low light, trying to do fast action .. can do if you have experience. But the probability is that it may disappoint, even if it's only just a little bit.


So the advice is(from a non Sony type that knows nothing of Sony cameras), turn off all fancy pants af features. you don't really need face detection(or do you?)
Single point af area, no grouping. if you have 199 af points and don't want to use all, your camera will usually have a menu item to select 198, or 99, or 11, or whatever. makes manual af point selection just a bit easier on the fly.
Continuous af mode is usually best for fast moving action.
if you have a feature where you can selectively focus using a rear button, try that (fantastically good for allowing you to control when focus is needed or not).

Aperture value will not(or should not) determine where af points should go .. the photographer really should!

While I did say turn off any fancy schmantzy af features, there may be one that could be helpful. On Nikon cameras it's called 3D tracking. It's almost an auto mode, but it's not.
What this one does is it tracks based on some colour parameters. if everything in the scene is brown .. it won't work.
But if you try to track a goalkeeper wearing a lime green top, against all the other players that are wearing blue or red .. it'll track the green topped goalie perfectly(mostly).
The only auto bit about it, is that once you activate the tracking, it then tries to track .. but you need to pick the initial spot to focus on.
if you cameras have anything like that, it could help. But you need to be in close or use along lens, as you want as little distraction in the scene as possible.

But again these so called features can actually be a liability in many situations, where consistency is needed.
As an example of that.
A few weeks back I was photographing my nephew playing soccer, he's a goalie, and wears this lime green top(see above). At one end all worked pretty much perfectly 90% of the time. every now and then the tracking may have gotten a bit lost, but if that was the game, you'd think 3D tracking was the best, most uber cool invention of all time.
But in the next half, the background was now more bushy(yep green!) so the colour tracking mode idea was a PITA, and the idiot that invented it should be made to wear Canon merchandise clothing! :D
Background completely stuffed it .. so it works and it doesn't .. face detect as you figured out generally does(n't)-ish ..

anyhow you get the idea.

You want certainty, learn to use it like a photographer would! .. consistently and under your control.

For exposure I always tend to use aperture priority. Make sure metering is as you expect, and once that's done, just shoot.
But focusing, basically manually full time. Do a quick search on AF-On, or back button focusing, and read some of the pro and cons .. check to see if your camera can be set up like that, and try it out too(if it can).

[bitsnpieces]

Hm... very good points indeed. Your rephrasing of my question was exactly what I was trying to ask - I was never good a phrasing.

I was in the 'AF phase' because I saw Sony's a99 advertisement and it's auto focus function, I was wondering how the a65 performed - which was one reason I used AF last night - really wanted to test what the a65 can do so I can know what my limits are, and how I can overcome them.

Regarding focusing on one point and just working with that intrigues me. I'll try for my next run. Because I can pick focus points to work on, constantly having to change them to work with a composition I'm thinking of though, can be a bit difficult when trying to get all the action down. Cropping will always be involved I guess to fine tune the composition too.
Midway last night, a reminding thought did come to me that 'no worries, you'll be cropping these anyways'. So focusing on one point, getting the clear shot, then cropping to the composition, just may be the way to go - I mean, I don't know much, and definitely have no idea how pros do it. lol
I've done the AF switch to MF before, mostly for fine tuning though, but something to keep in mind again. Thank you :)

The Sony a65 has the 3D tracking you also mentioned, don't remember what Sony labels it though, but I avoided that knowing it's limitations - our multicultural hall isn't very colourful, nor most the player's attire, it'd struggle.

Definitely some improvements in the way I photograph, and some good insights here.

But still, one question remains, for when I do want to use AF, and more specifically, the face detection (because during last nights test, it'd detect the face, but not focus there, unless the face was in one of the AF points, then sometimes it'd work), I'm hoping to get some light shed on this.
Like point & shoots, they'll see the face and focus on the face, but last night, it sees the face, but didn't focus on as I thought it should. Was just wondering if I had messed up a setting or something? Or maybe I wasn't in the right conditions for it to work properly? Don't know.
I just think it'll be useful to have if taking a group photo. Sure, manually focusing works too, but as mentioned, having AF has it's bonuses.

[ricktas]

Some more.

AF systems work on contrast. Try getting your camera to focus on a pure blue sky (no clouds), it will 'hunt' for focus. Then try and get your camera to focus on a nice fluffy cloud and see what happens. AF systems work by 'seeing' the contrast in what they are aimed at, and thus anything that is one plain colour/brightness will cause the AF system to go 'huh..what am I looking for' and it will 'hunt'. Now consider what you were trying to focus on at this sports night. Did you try the players faces or their plain blue tops (no idea if their tops were plain or blue, but you get the idea). So to add more to the King's ^ post, use the AF system knowing how it works and aim at something it will 'lock onto'.

Face Detection. You mention point and shoots, so now lets consider that. When you use a P&S with face detection, think about how long the camera takes to firstly recognise there are faces and then focus on them. Now swap to your sport shoot. You probably wanted to get the action as it happens, but if you have face detection turned on, your camera is doing the same as the P&S, it is running the face detection systems and trying to find faces, then when it does, focus on them, but this time your subjects are moving around. So by the time the camera software goes, Oh ah face, lets focus on it... that face has turned around, moved away, click..miss! Turn face detection off. Only use it for posed portrait stuff, if you want to. Photos of people have been taken well for nearly 200 years now and we haven't needed face detection..it's a bit gimicky.

Aperture, this has something that most beginners do not realise. As you have said about being wide, the smaller the number (you mention f5) the narrower the depth of field. So shoot something at f5 and then at f22 and more in front and behind your subject will be in focus with f22. And this is the basics of how aperture works. But most beginners miss one bit. Distance to subject. The distance from the camera to the subject also affects how much is in focus.

On the "AP Extras" menu (sub menu at top of site) is this little fellow: http://www.ausphotography.net.au/aptools/ap_dof.html

Have a go at it. Set it up, then change the aperture and see how the depth of field changes. But then leave the aperture etc alone and change the distance to subject and see what happens. Do not get to concerned about what is in focus when shooting at f5 and the subject is 25 metres away..as the DOF calculator will show you. In fact most photographers want the opposite, they want a narrow depth of field over great distances, thus another reason we see f2.8, f1.8 and f1.4 lenses in existence.

So turn off all the gadgetry your camera has, choose a good ISO, Aperture, Shutter Speed, use focus tracking to follow your moving sports people, or if they are not moving that much, use single point AF, and have fun learning how to use the camera, without relying on all the stuff they add to cameras these days. The very basics of photography are often the best way to get good results.

And finally there is one other thing. Your lens might not be perfect (no lens is) and it might actually be front or back focusing, that is focusing in front of the subject or behind the subject. There are ways to fix this, but first you need to test to see if your lens is focusing at the point you want it to or somewhere else.

http://www.ausphotography.net.au/forum/showthread.php?109346-Focus-Test-Charts&highlight=focus+charts

http://www.ausphotography.net.au/forum/showthread.php?96874-Lens-Calibration-Tools&highlight=focus+charts

[wmvaux]

I don't know your camera and the various focus setups, but I've had success shooting sports in manual at 1/1000 at f6.3, Auto ISO and using a 9 point area focus. In the past I made the mistake of using too great an area for focus points, either 51 or one of the other dynamic focus settings which covers most of the view finder. The problem with this is if there are other things moving that might take the focus away from the main subject.
I haven't taken photos of basketball or indoor sports, so no doubt the ISO would need to go up quite a bit and it should be more stable than outside, so you may be fine just fixing the ISO, just check your histogram after shooting in different directions to see if it changes, if it does, then I would go with Auto ISO. Regarding the f stop having a medium aperture will give you better quality images and increases the depth of field over your f5 setting.

It can be trial and error, different conditions mean different settings are required. Good luck and keep shooting :th3:

[arthurking83]

Ah! right .. hopefully you're not getting too confused now as others have also entered the thread.

Rick mentions aperture and DOF, and focus distance. Everything he mentions makes sense(for you to try), but hopefully hasn't confused you too much either.
You haven't mentioned which lens you used(you have two listed in your sig!) and at roughly what distance from the action you were.

Being netball and volleyball, were you right at the edge of the court/s(ie. close in), or 'up in the rafters' with all the other spectators(ie. far out)?
These specifics can make all the difference. For better focusing(with what you have) get closer!
This is one way you can make the a65 focus better.


......
Could it be my aperture is too wide, so it doesn't have the breathing room to focus on multiple points? I was using f5 throughout the shots, hoping there would be enough to focus with, but not too small so that I can still use 1/160 and ISO 3200.......

Because you said this, I'm assuming you assume that as you vary the aperture, the camera's af system may vary in the way it works, or how well it works.
No( .. but yes!)
If you used your 28-75/2.8 lens at close range and stopped down to f/22, the focus system of the lil a65 would be neither better, nor worse off .. it'll work exactly the same.
This is because of the way cameras work nowadays.
With your normal lenses, the largest aperture is always maintained during stills photography, unless there is a way to vary it on your camera.
So, if you set f/22 on the lens, during af and during composition, just prior to exposing the shot, the cameras forces the lens to stay wide open.
At the precise moment of exposure(a nano sec before) the lens is stopped down to f/22 for the shot.
So like Rick alluded, with the change in aperture you dial in, you will see a DOF change(or maybe not) but only in the resultant image .. not during the composition phase, unless there is a way for you to specifically set this(some cameras have this ability).
Look/search for Depth Of Field Preview(DOFP) button operation, or full manual aperture control.

Now vmvaux mentions that they prefer to shoot at f/6.3 which works well for them. And their comment about the extra DOF is also relevant and accurate.
But again, for your focusing problem, and in a different way f/6.3 could introduce an issue for you.
Because you have the 18-250mm f/3.5-6.3 lens listed in your kit, by using this lens at the longer end of the zoom range, could result in less accurate/slower/lethargic/indecisive/etc focusing from the camera.

Where before with the f/2.8 lens stopping down didn't affect af, mounting this lens and zooming in even to a medium degree, actually can.
The actual physical maximum aperture size of the lens has an impact on the cameras ability to focus efficiently.

So, the nett result could have been something along these lines:
You had the 18-250 lens mounted set to 200mm and trying to use face detection(or whatever .. maybe just group af area).
Because at f=200mm the aperture value is a teeny f/6.3 and that is now how small your aperture actually is, it may have been a bit dark for it to be sure it was a face.
A human analogy is like looking through the cardboard roll of one of those kitchen paper comes rolled up on. The entrance pupil of the light is much smaller and the af system which has to see into this light will now struggle.
Whether this is a little or a lot is massively dependent on the camera AND lens combo. Some slow lenses focus quick smart, some fast lenses focus quite averagely slow. So if you use a 'slower lens' you may experience af difficulties .. especially on consumer spec bodies.


Just so that you understand completely too, my use of the aperture values and implementations is for the purpose of autofocusing, and not as Rick and vmaux have done and used them for DOF in photos.
Hopefully from this you can see why so many of us prefer the company of fast lenses. Not only for the lower ISO, or shallow DOF, but that they generally focus a bit easier/better(especially at the longer end).

ps. this info may eventually become relevant to you because of your adapter/classic lens interest thread too as once you use non standard lenses, you may be back to full manual control, which is a PITA sometimes.

[bitsnpieces]

I don't know your camera and the various focus setups, but I've had success shooting sports in manual at 1/1000 at f6.3, Auto ISO and using a 9 point area focus. In the past I made the mistake of using too great an area for focus points, either 51 or one of the other dynamic focus settings which covers most of the view finder. The problem with this is if there are other things moving that might take the focus away from the main subject.

That may have been the problem - too wide of a spectrum for the camera to work with, how would it know what it needs to focus on right? :)



Face Detection. You mention point and shoots, so now lets consider that. When you use a P&S with face detection, think about how long the camera takes to firstly recognise there are faces and then focus on them. Now swap to your sport shoot. You probably wanted to get the action as it happens, but if you have face detection turned on, your camera is doing the same as the P&S, it is running the face detection systems and trying to find faces, then when it does, focus on them, but this time your subjects are moving around. So by the time the camera software goes, Oh ah face, lets focus on it... that face has turned around, moved away, click..miss!

I guess that probably goes in line of what wmvaux said also, the camera struggling to focus on too many points. Will see what happens with detection off.


In fact most photographers want the opposite, they want a narrow depth of field (http://www.ausphotography.net.au/forum/showlibrary.php?title=New_To_Photography:Depth_of_Field_and_Aperture_-_why_is_it_so) over great distances, thus another reason we see f2.8, f1.8 and f1.4 lenses (http://www.ausphotography.net.au/forum/showlibrary.php?title=New_To_Photography:Using_different_lenses) in existence.

That's interesting indeed - I'm guessing it's so the shot looks clean right? Up close, clean shot of their target, things else blurred? Understandable. I presumed it would bit a little harder to focus is all, because of the deeper depth of field, so I thought using f5 would be okay, not too deep, and incase I miss the focus (I had pretty much anticipated this would happen already lol), I can still get a decent amount of focus going - breathing room.

But as mentioned, last night was to really see what the a65 could do, throw on its fancy features and let it rip, managing 1700 pictures with 50% power left, that's a lot of photos... surprised my inexperienced self. :P


But regarding face detection, thus, still unanswered and hoping to get a clearer understanding.
So understandable in sports, probably not efficient, but in a portrait or group photo, this is the purpose of it pretty much. True, we didn't need it back in the days, but it's a nice feature, and a feature I don't know how to make work on the a65 yet, thus my questioning here - am I using the wrong setting somewhere?

Because I did take a few group photos, static poses, people sitting down, smiling, no problems. Boxes appear, but when I click to focus, it focuses to the centre AF point, which is on their chest or knee or wherever it may be at the time. Maybe a little off centre too, but never the face. I used Wide AF area because I read that was required to give the camera room to scan the image and focus.
My only ideas are, the face didn't line up on one of the AF points, which I think is ludicrous if this is a requirement (so definitely won't be needed face detection), there wasn't enough light/contrast to focus with (but things were well lit and the boxes did appear after all - it knows the face is there), the aperture was too wide (so with the big depth of field, it would struggle recognising multiple faces across the plain, I guess?), and who knows what else. Maybe I had a wrong setting somewhere and didn't realise that, meh?

Reason I ask about this is more so because I had my Sony Cyber-shot DSC WX300, face detection worked very quickly on that. The boxes appears, it knows the face is there, I focus (almost instantly focuses the face), and snap. It didn't snap at a random other focus point where you get the multiple squares that the point & shoots focus on, it was dead on the face most of the time, sometimes if the subject moved and it was able to keep up.

I guess that's only my real beef about the a65 at the moment. It's not that it's a must have tool, but it's a nice function to have if I ever want to use it, just like working on manual focus is fine, but auto is nice to have and makes life easier (in reference to my post about finding another lens to use here (http://www.ausphotography.net.au/forum/showthread.php?131647-Adapters-to-A-Mount))

So yeah, just trying to make sure I know how to make use of all the features.

I did notice though that the 28-75mm felt to focus slow a little... The auto rotating wasn't hugely quick. Catching was easy, just the speed to focus - I guess it's trying to move quietly? Or I could be over expecting.

[kevin301]

.
.
.


But regarding face detection, thus, still unanswered and hoping to get a clearer understanding.
So understandable in sports, probably not efficient, but in a portrait or group photo, this is the purpose of it pretty much. True, we didn't need it back in the days, but it's a nice feature, and a feature I don't know how to make work on the a65 yet, thus my questioning here - am I using the wrong setting somewhere?

Because I did take a few group photos, static poses, people sitting down, smiling, no problems. Boxes appear, but when I click to focus, it focuses to the centre AF point, which is on their chest or knee or wherever it may be at the time. Maybe a little off centre too, but never the face. I used Wide AF area because I read that was required to give the camera room to scan the image and focus.
My only ideas are, the face didn't line up on one of the AF points, which I think is ludicrous if this is a requirement (so definitely won't be needed face detection), there wasn't enough light/contrast to focus with (but things were well lit and the boxes did appear after all - it knows the face is there), the aperture was too wide (so with the big depth of field, it would struggle recognising multiple faces across the plain, I guess?), and who knows what else. Maybe I had a wrong setting somewhere and didn't realise that, meh?

.
.
.
.

I am not a Sony user, so I'm not sure if my answer is correct.
Could it be that you have a setting for "'face detection on/off" and also one for the auto-focus modes (single point, wide area, 3d tracking, etc).

So for example if you switched on face detection but have the AF mode set to single point in the center, the camera cannot over-ride your selection and thus is restricted to focusing in the center only.

edit: I just downloaded the a65 user manual to browse through it, and there is a reference made about "compose to overlay the face detection frame and the AF area..." which suggests that there are 2 factors at play here.

One thing to note is that with most cameras and lenses nowadays, the lens is set to the largest possible aperture until the split-second before the image is recorded, that the aperture is closed down to your set value. This allows the maximum amount of light into the camera for focusing purposes.

[ameerat42]

Bitsn... Have moved your thread to Shooting Help.
Hope all goes well.
Am.

[bitsnpieces]

So for example if you switched on face detection but have the AF mode set to single point in the center, the camera cannot over-ride your selection and thus is restricted to focusing in the center only.

Yeah, I put my guy on wide area so that there would be no restriction on where the focus can occur - it just seemed to prefer the centre area


edit: I just downloaded the a65 user manual to browse through it, and there is a reference made about "compose to overlay the face detection frame and the AF area..." which suggests that there are 2 factors at play here.

Based on that, I feel that may be in line with my guess that the faces, though detected, need to be within one of the AF points (the AF area), as to catch that face more accurately and focus from there.

I'll see if I can do a mock image to put up and to help illustrate the issue, and hopefully if this is the case, it may help other Sony users out there understand the function more.
Though still perplexes me how the point & shoot can focus on faces better - I'm simply guessing it has a wider acceptable AF area, so it works better.



@ameerat42 - thank you. No problems at all :)

[kevin301]

Yeah, I put my guy on wide area so that there would be no restriction on where the focus can occur - it just seemed to prefer the centre area



Based on that, I feel that may be in line with my guess that the faces, though detected, need to be within one of the AF points (the AF area), as to catch that face more accurately and focus from there.

I'll see if I can do a mock image to put up and to help illustrate the issue, and hopefully if this is the case, it may help other Sony users out there understand the function more.
Though still perplexes me how the point & shoot can focus on faces better - I'm simply guessing it has a wider acceptable AF area, so it works better.



@ameerat42 - thank you. No problems at all :)

I suppose the Depth of Field being much deeper with a PnS makes it harder to notice if something is out of focus :)

[bitsnpieces]

Okay, so based on things discussed here and through experience, and theories, I've made some mock images to help illustrate the issue, and hopefully if correct, can help people with Sony Alphas. If not, then hey, we're all learning. Hopefully a more experience Sony Alpha user who may know more about this can help also.

Again, this is all theory, though I am confident about the WX300 because I never really had trouble using Face Detection on that thing.

So, here's a photo taken at the airport before my sister went off to Thailand to teach English on my Sony Xperia Z, from left to right, myself, mother, sister, brother, father

This is all based on using Wide AF Area
https://farm8.staticflickr.com/7393/13973766049_741355a439_s.jpg (https://flic.kr/p/nhPcNR)00 Photo (https://flic.kr/p/nhPcNR) by bnpbitsnpieces (https://www.flickr.com/people/123455861@N06/), on Flickr

So as we know, there's Face Detection, so it'll look something like this:
https://farm8.staticflickr.com/7449/14160441905_8e9e2f853c_s.jpg (https://flic.kr/p/nziY36)01 Face Detected (https://flic.kr/p/nziY36) by bnpbitsnpieces (https://www.flickr.com/people/123455861@N06/), on Flickr

This is our goal right? Detect, focus, and snap, on all faces
https://farm3.staticflickr.com/2938/13973804248_d95032b9fc_s.jpg (https://flic.kr/p/nhPpas)02 Face Focuses (https://flic.kr/p/nhPpas) by bnpbitsnpieces (https://www.flickr.com/people/123455861@N06/), on Flickr

Now here's Sony's AF Points:
https://farm8.staticflickr.com/7350/14157649152_22d741cc37_s.jpg (https://flic.kr/p/nz4DRb)03 AF Points (https://flic.kr/p/nz4DRb) by bnpbitsnpieces (https://www.flickr.com/people/123455861@N06/), on Flickr

So this is where the AF Points line up to the Face Detection
https://farm8.staticflickr.com/7400/14160442655_b1187ca9bb_s.jpg (https://flic.kr/p/nziYg2)04 Focal Alignment (https://flic.kr/p/nziYg2) by bnpbitsnpieces (https://www.flickr.com/people/123455861@N06/), on Flickr

Maybe, just maybe, it'll focus on my sister - make the little stretch
https://farm3.staticflickr.com/2895/14157094131_47201a7c84_s.jpg (https://flic.kr/p/nz1NRR)05 Maybe Focus (https://flic.kr/p/nz1NRR) by bnpbitsnpieces (https://www.flickr.com/people/123455861@N06/), on Flickr

Most likely, it'll focus in the centre, or one of the AF Points around there - the camera likes the centre
https://farm3.staticflickr.com/2910/14157649792_4a5640aed4_s.jpg (https://flic.kr/p/nz4E3d)06 Most Likely Focus (https://flic.kr/p/nz4E3d) by bnpbitsnpieces (https://www.flickr.com/people/123455861@N06/), on Flickr

What if I had a closer crop? The AF Points now line up on some of the faces detected
https://farm6.staticflickr.com/5567/14157092641_8a6e950432_s.jpg (https://flic.kr/p/nz1Nqa)07 What If (https://flic.kr/p/nz1Nqa) by bnpbitsnpieces (https://www.flickr.com/people/123455861@N06/), on Flickr

Most definitely, my sister should now be in focus, and my father, high probability, but no other faces fall into the AF Points, so they don't get focused - This is what I believe my problem was when taking some of the group photos during the sports night
https://farm3.staticflickr.com/2935/14160535554_2f9a9b88ce_s.jpg (https://flic.kr/p/nzjrSJ)08 Possible (https://flic.kr/p/nzjrSJ) by bnpbitsnpieces (https://www.flickr.com/people/123455861@N06/), on Flickr

What if it was a closer crop? AF Points are reaching more faces
https://farm8.staticflickr.com/7304/14160535424_c944bfe5dd_s.jpg (https://flic.kr/p/nzjrQu)09 More Coverage (https://flic.kr/p/nzjrQu) by bnpbitsnpieces (https://www.flickr.com/people/123455861@N06/), on Flickr

In theory, my mother, sister, and brother should all be in focus. My father, again, very high probability (I didn't paint it green this time)
https://farm8.staticflickr.com/7322/14160440985_0ed2268d92_s.jpg (https://flic.kr/p/nziXLe)10 Result (https://flic.kr/p/nziXLe) by bnpbitsnpieces (https://www.flickr.com/people/123455861@N06/), on Flickr

That's my theory on how the Face Detection works on the Alpha series using Wide AF Area and not Object Tracking. I will soon need to set up some things so I can test this theory out some how

So here's why I believe the WX300 worked so well - again, I don't know where the AF Points are, but I do believe I have it correct. When using the camera, I felt the AF Area was very wide.

This is a representation of what I believe WX300's AF Points are at, in coordination with the Face Detection
https://farm3.staticflickr.com/2919/13973802718_f696b0f9bb_s.jpg (https://flic.kr/p/nhPoH5)11 WX300 (https://flic.kr/p/nhPoH5) by bnpbitsnpieces (https://www.flickr.com/people/123455861@N06/), on Flickr

Now, because the faces are all within the field of the AF Points, they all get focused very easily
https://farm8.staticflickr.com/7323/14180514073_2dc354ec6b_s.jpg (https://flic.kr/p/nB5QMX)12 Perfect (https://flic.kr/p/nB5QMX) by bnpbitsnpieces (https://www.flickr.com/people/123455861@N06/), on Flickr

Of course, Object Tracking is a feature on both cameras, so you can easily pick a spot you want to focus on, like my hands, the luggage, an individual face, my father's shoes. However, what I've done is put a red box on a face on the wall. Whether the camera detects it or not is one story, but it definitely will not focus there - it's out of the AF Area and the AF Points, and the only way to track it is use Object Tracking. But, if I remember correctly, Object Tracking doesn't go all the way to the edge of the screen, only very close, like half way in the AF Point, and half way out, thus if I focused on the wall painting, recompose, it'll fall out of focus again. I don't know how close to the edge Object Tracking can go on the a65 yet.
https://farm8.staticflickr.com/7380/13973917309_8d8301b50d_s.jpg (https://flic.kr/p/nhPYLM)13 Object Tracking (https://flic.kr/p/nhPYLM) by bnpbitsnpieces (https://www.flickr.com/people/123455861@N06/), on Flickr

So, what do you all think? It's a lot, I know... Sorry lol

[kevin301]

Actually from the way they are standing in a same line, practically equidistant from your camera, as long as 1 is in focus, with a reasonable depth of field all of them would also be in focus. If they were standing 1 in front of the other then you would be possibly having issues.

Usually face detection is working with contrast detect AF (CDAF), where the camera is literally analyzing the live view image whilst adjusting the focus back and forth.
The way most DSLRs focus is with Phase Detect AF (PDAF) using the AF sensors. This is a much faster method of autofocus, and all it requires is something with contrast for it to determine focus accuracy. With PDAF, you don't have to concern yourself with face detection.

[arthurking83]

I was also going to mention that face detect may use CD focusing as it detects and focuses.
My only reservation about that tho is that the Sony uses an SLT design and hence has full time PD focusing available to it.
So, if it's true what bits says about the face detect tracking being limited by the available focus area points, then this will be due to a system where the face detect system is not adjusting focus using contrast detect.

Contrast detection of course is not limited to the af areas and can be deployed across any part of the sensor .. right into the very corner if required. CD focusing uses the image on the sensor as a way to adjust for focus.

Phase detect on sensor is a rare beast, and AFAIK, only Sony's newest sensors, ie. released after the a65, and a very few others(Nikon 1) may have it.

Traditional PDAF has been based on the separate af system model(which is what the a65 uses) .. and PDAF can only use those sensor areas.

The WX300 will only have CDAF focusing, and (most likely) no PDAF system.
(note I don't know this for sure, just educated gueswork, based on the type of camera the WX is!)

You can't really compare the focusing ability of the Wx against the a65, even tho the two cameras are still focusing on the same subjects.
the a65 is at a disadvantage in a few areas, the main one being that it has an inherently shallower DOF to deal with, so AF has to be a few orders of magnitude higher in terms of capability to the Wx. Also, the lenses that the can be mounted to the a65 vary so wildly in ability and FOV, that it's impossible to separate and determine if the lens is causing any issues, or is it just the camera.
So while the WX enjoys the massive advantage of natural DOF, so much so that it probably doesn't even need to focus at all and it may still looked focus(focal length is probably something like 5-6mm, so do the DOF calcs to understand what/why) .. but of course the major disadvantage will be ability to maintain focus accuracy whilst tracking and low light fast paced situations.

Turn off the lights, get them all to hold candles and see if the WX will still track faces!


The question with your a65 is(or should now be) .. is the focus tracking during liveview(rear LCD viewing) mode, based on phase detect, or contrast detect.
Sony claim full time PD for their SLT cameras, which makes sense going by the design of the camera with it's fixed pellicle mirror.
It almost certainly has contrast detection for AF too, very few movie enabled cameras don't have it!

A subsequent question would be, if can you configure the camera to use CDAF or PDAF during rear LCD viewing mode.

[bitsnpieces]

True, and that's one idea I had, was that my DOF was too shallow, thus the AF wasn't able to catch the Face correctly, but at the same time, if there is priority to focus on the face, the camera would have moved its DOF/focus to the face, and not an AF Point as illustrated.

The WX300 I had had no problems detecting, focusing, and capturing faces at different heights and distances (people standing behind someone else), as long as they were within the area of focus.

So here's a couple more examples of groups I took that night which brought up this conundrum:

First one - they seem pretty even, there really isn't much of who's in front and who's behind, except the children, but even so, very very little and I was at F6.3, should be plenty of room to focus on multiple faces. Now, I don't remember if any boxes even appeared to say the camera has detected a face (so already, you can see something is wrong), but even if there was, the green marks where the camera focused (I don't remember which of the two, but it was one of them)
https://farm8.staticflickr.com/7307/13980035958_fb05acea01_s.jpg (https://flic.kr/p/ninkCJ)Focus Here (https://flic.kr/p/ninkCJ) by bnpbitsnpieces (https://www.flickr.com/people/123455861@N06/), on Flickr

Based on my theory, the lady in the middle holding the kid, both their faces should have been focused (if their faces were detected) because their faces are inside one of the AF Points, heck, even multiple. So like I said, I don't remember if face boxes came up or not to begin with.

But here's my next example, which did have face boxes appear, however, when I went to focus, it didn't focus on the face, only the middle.
https://farm8.staticflickr.com/7427/14186761383_d8aa9de5ce_s.jpg (https://flic.kr/p/nBCRUc)Was Here (https://flic.kr/p/nBCRUc) by bnpbitsnpieces (https://www.flickr.com/people/123455861@N06/), on Flickr

Yes, there's contrasting colours in the middle, easy to pick and focus, and none of the faces were within the AF Points (which is in accordance to my belief as illustrated previously) but here's the question; isn't the point of Face Detection is to put priority on the face to focus there?

So this has left me with the theory explained above, they need to be within one of the AF Points because when using the Wide AF Area, the camera will base it's focusing on one of the points. It will looking for contrasting things in that point, and the one it thinks is most important, it'll focus there. So the theory is because the face isn't in one of the AF Points, the camera ignores it because it's only looking at the 15 AF Points. If a Face was there and the camera can pick up the difference in contrast as well, bang, face detected, confirmed, focused, shot. Because the only time I noticed a face actually focusing from the Face Detection feature was when it was within an AF Point.

Which leads to this problem - Do I need to register faces first? If I had registered a face, would Face Detect override the importance of the AF Points and the camera will focus on the face, because I'm assuming by 'registering a face', it must put priority on that face. So the effect will be like Object Tracking.

Which is why Object Tracking works, you've made a choice to follow something outside of the standard 15 points, so it's going to follow that choice - You've chosen something of higher importance, is this the effect when registering faces?
If so, I find this very redundant and a problem - isn't the whole idea of detecting faces in general is to put priority on them, and if you want to put priority on certain faces like the wedding couple in the group, then registering helps keep that in check? Least that's what I understand.

True, if I wanted to, I can use object tracking, but that puts the focus only on one thing, the purpose of Face Detect is to make sure all faces are in focus right? So if you're DOF is too shallow, the camera presumably should bring up the aperture to get a wider DOF to cover all the faces right? I was using shutter priority, so it shouldn't have restricted the aperture from changing if it felt f6.3 was too shallow to catch the face in front and behind, so it'll use, f8 for example (which is one flaw I found in the WX300 if I remember correctly - not all faces detected and focused were actually in focus - possible hardware or simply too slow a shutter speed and I shook a bit too much?).

So there's my spill.

Again, I do plan on testing out this theory so it can be confirmed, but from experience so far (and based on memory), it seems this way.

And true, we can't compare the WX300 to the a65 like I am, but this is mostly to help me (and maybe other Alpha users) understand the Face Detection function on the camera, and possibly even other aspects such as how focusing works as arthurking83 mentioned.

I don't believe there's an option with regards to the different methods of AF, it's automatic, so whether it's PD or CD is a good question.
But I feel also that it wouldn't simply be about CD or PD right? Because as you said, CD should be able to focus all over the screen (I wouldn't be surprised if Object Tracking used CDAF to do so). I don't know how PD works, but I assume it's the same as CD, just faster because the mirror is constantly there, allowing the constant contrast checking for focus.

Which comes to the question I put out be more in relation to how Sony implemented the Face Detection function in the camera - did they mess something up? Is there a hardware limitation because of how they use PDAF and CDAF? I feel the problem is more of application of technology - but I could be wrong. Maybe it is as simple as a few setting changes like aperture, or lens, or whatnot? *shrugs*

[kevin301]

PDAF would be using the specific AF sensors that are present on the camera.

The number and coverage area of these sensors increase in relation to the price of the product :)
Because of Sony's SLT technology, they are in a fairly unique situation to offer both PDAF (thru AF sensors) and CDAF (through the image sensor itself) at the same time. This opens up the possibility of 'clashing' though.
On my Nikon, I have face detection but only in LiveView (CDAF), so there's no issue about getting mixed up.

My suggestion would be to forget about face detection because it's too gimmicky ;)

[bitsnpieces]

Yeah, I've taken the Face Detection on their DSLRs as a gimmick at this rate, but it'll just be nice to understand how it all works :)

Hopefully this long thread has given some insight or something to something out there. :P

[bitsnpieces]

I believe I have found the answer to this problem, which confirms my theory:

http://www.dpreview.com/forums/thread/3140315
It's the 5th post, post by Kilrah in reply to William Porter where he says the AF Point has to be in the face too.

Which two posts afterwards, he explains his test and how Face Detection doesn't override user choice in AF Points.

So I guess this clarifies 2 things:
Using Wide AF Area only works if you set an AF Point onto the face, as I have illustrated.
And Face Detection basically only works in Wide AF Area because the other AF modes are user choice (obviously if you've chosen a spot where the face is, it'll focus anyways, but the automation of it isn't really there).

So there you have it, Face Detection works, but only under certain conditions, thus in a DSLR, it's a gimmick (though we've all established that already). In a Point & Shoot, it's magical - go for it.

[kevin301]

I believe I have found the answer to this problem, which confirms my theory:

http://www.dpreview.com/forums/thread/3140315
It's the 5th post, post by Kilrah in reply to William Porter where he says the AF Point has to be in the face too.

Which two posts afterwards, he explains his test and how Face Detection doesn't override user choice in AF Points.

So I guess this clarifies 2 things:
Using Wide AF Area only works if you set an AF Point onto the face, as I have illustrated.
And Face Detection basically only works in Wide AF Area because the other AF modes are user choice (obviously if you've chosen a spot where the face is, it'll focus anyways, but the automation of it isn't really there).

So there you have it, Face Detection works, but only under certain conditions, thus in a DSLR, it's a gimmick (though we've all established that already). In a Point & Shoot, it's magical - go for it.

well I don't know about "magical" :)
ever since I got my first DSLR I have hardly had the chance to use a PnS, except on the rare occasions where people ask me to help take their photo. I'd still choose the PDAF system any day! So much faster, accurate, and also predictable.

[bitsnpieces]

True... But P&S still work great, and their face detect work well, least on the Sony. :)

Now to look into the information threads to see where I can add this information to :)

Interesting note also is how small the PDAF area actually is, and I guess it's pretty evident when you look at where all the AF Points are at:
http://aportal.sony-asia.com/learn/article/about-phase-contrast-and-hybrid-autofocus

In the Hybrid diagram, you can see that the PDAF, though very accurate and effective in what it does, is very constraint. CDAF gives more coverage though - wish the Sony made use of hybrid so give us absolute AF depending on the situation (don't know if the a77/99 has hybrid or now, but definitely my a65 doesn't :()

[arthurking83]

.... (don't know if the a77/99 has hybrid or now, but definitely my a65 doesn't :()

AF ability and features is an important feature distinction set that manufacturers use to separate lower end models from higher end models.

And just to be clear on this face detection issue: The problem of face detect and the af points happens irrespective if you are using the viewfinder, or if you are using the rear LCD screen to shoot?

I find it strange if the issue happens in rear lcd shooting mode.
When using the vf, it makes sense that face detect(if it works) would be tied to the AF point only if face detect via phase detect were possible.
But it may seem that it isn't.

The way PDAF works, is that the PD sensor basically splits the in coming light beam of the object into two different phases(think of it in a similar way to those red/blue 3D glasses.
If you know how a rangefinder focusing system works, then you get the idea of what a PDAF system is doing in a basic sense.
When the two phases are in sync, ie focused .. the PDAF systems recognises it as in focus.

A CDAF system looks for contrast between adjacent pixels. At it's most basic image in black and white line(Collingwood jumper) pattern. When in focus the difference between the black and white lines is sharp and obvious. CDAF sees this, but it takes time to see it clearly. The CDAF system has to focus defocus in small steps to see where the best sharply rendered difference lies.

Where PDAF deliberately splits the image and then uses that split to recombine it again into a focused image, CD sort of has to construct the image it needs to see from the start.
There's a ton more programming smartiness in CDAF, and hence why it also recognised faces .. and so many types of faces.

I don't know for sure, but I don't think PDAF can have the sort of programming to recognise specific objects(such as faces, or features).
it's prgrammability seems to be limited to recognising only colour, or patterns of colour(subject tracking).
It doesn't know the subject being tracked. It just recognises the differences in colour between it and it's surrounding area.


I'm just surprised that in rear LCD shooting mode, the a65 wouldn't have full time CDAF and hence full time access to face detect.

On the D800, when in liveview mode(rear LCD shooting), the PDAF points aren't active. The CDAF system is totally independent of the PDAF system. There are no AF points as such(except that you can snap to the centre with one action if need be) .. the entire sensor area is an AF point.
But CDAF is slow as hell and you wouldn't use it for every day focusing.
Once it's focused and tracking faces tho, and if the lens is fast enough to hold focus, it can be useful if needed.

In fact this is the first time I've actually turned face detect on and tried to see it in action.

[bitsnpieces]

That's a very interesting point posted up there, that when using liveview, it should go into CDAF mode - based on Sony's little diagram, it would seem that should be the way it works (given that it has hybrid AF) - I'm going to double check my settings to make sure I haven't set something that is blocking this.

Or I could check Google to see if a65 has hybrid AF or not :P Who knows... Maybe one feature Sony took out from the a65 to become the little brother to the a77 is hybrid AF...



Okay, I'm calling it, the a65 has no CDAF, even Object tracking didn't focus unless it was on an AF Point.
Basically, everything relies on the AF Points - if what you want to focus on, face, object tracking, etc, the initial focus won't happen unless it's on an AF Point, and the camera will just focus in the centre or whatever it likes.

Face Detection priority works the same way. You put priority on a face, but the AF Points line up to a different face, it'll face the other face - so priority only works if there's an AF Point on it also, then out of the two faces with AF Points, it'll pick the priority face.

Using AF Centre or AF Spot (or whatever it's called when you pick an AF point) will override Face Detection. It will also focus in the centre or the AF Point chosen, never on faces, pretty much.

Also, I could not get Face Detect to focus on multiple faces at the same time, it always focused on one only. When I put object tracking on the far right, then went back to get both faces on AF Points and the object tracking out of one, then the camera focused both faces sometimes, most of the time still only focused on one face. And now because the object tracking is left dangling far right, it easily loses focus of what it's tracking and turns off.

Well, I think this concludes my theory and also verifies that the a65 isn't using CDAF, and if it does have it, I have absolutely no clue how to access it - menus all seem fine.

[arthurking83]

There may be a subtle difference between object tracking and CDAF on the little Sony.

I find it weird that Sony wouldn't have CDAF on a camera that is heavily biased towards video.
PDAF is either rare or non existent in video modes, and all cameras I know of use CDAF.

So ...

maybe you need to be in a normal non tracking type AF mode(well maybe face detect or something) AND you may need to use one of the automated scene modes to get the camera to use CDAF or something like that.

First off set it to video mode to make sure that it won't use the PDAF sensor points.
Then once you can confirm that video mode does use only CDAF, then you know that you have a CDAF only mode.
Once you know this, try setting the camera to a scene mode(Portrait would be the most obvious choice) and see if you can get CDAF only whilst trying to track and focus using liveview mode(not the evf mode).

Just had a read of DPR's review of the a65, and it looks like the video mode does rely on the PDAF system, as does tracking.(gives info about face detect mode tho).
DPR says that for video mode and AF, you need to set the camera to Program AE mode. If you try M, A or S modes, you don't get AF in video mode as the limitation is that the aperture needs to be set to f/5.6 or larger. This is always a limitation of PDAF sensors at the cheap end of the market, higher end PDAF cameras work down to f/8 nowadays.

So it seems that the a65's focus system is limited heavily in many ways.
DPR also goes on to summarize that while the focus is generally good for most situations, fast paced action is a weak point for the af system.

[bitsnpieces]

My next test was to try one of the modes, and forgot to do it today when I had the chance *dough*

But yes, that is something to test to see if CDAF is condition or not.

I'll try and video too - I believe I should have time later today

[kevin301]

The interesting thing about the Sony SLT range is that because of the fixed translucent mirror, there is always light falling onto the AF sensors during LiveView. Thus it presents a rather unique scenario of being able to utilize the AF sensors for PDAF, and the image sensor for CDAF. With typical DSLRs, you either get 1 or the other :)

As Arthur mentioned, the AF sensors for PDAF work on beam splitting and analysis to determine the focus error. This system would not be able to recognize faces.

Therefore I reiterate my stance that because of the different nature of PDAF and CDAF, it doesn't make sense for a ("gimmicky") feature like face detection to be able to work with both :)

Perhaps the camera is simply going "Hey, here's a face I detected. Please align it with one of the AF points so I can use PDAF for more accurate focusing"

[bitsnpieces]

I believe you're right Kevin - there's just no way to get anything to be focused unless they're on one of the AF Points.
I've tried the portrait modes and movie mode - gotta be in the square.

So it's unfortunate - though using PDAF which works fast, there are still many limits such as a small area of focus, and if anything ever happens outside of it, or for convenience, it doesn't cut it.

Well, I guess you just work with what you have, but the other disappointing factor is that it only focuses 1 face at a time, rather than multiple, even if there are multiple faces lined up onto an AF point... :S

Well, case closed.

[arthurking83]

so when the face detect system is on, and it sees the faces off the PDAF sensor area, does it recognise the faces?

The way I've seen face detect working is that it finds a face(or faces) and then follows them(phone, DSLR and P&S... they all do this)
If it recognises the faces, the colour of the squares determine if it's being used for focus.

So if you have multiple faces detected, and some of the squares are yellow or red, and others are green, then it's obvious that the green ones are the ones being used for focus.

CDAF is much slower to operate than PDAF, so it may not be immediately obvious that AF is working. The colour of the focus highlight is a better indicator.
And don't assume that becasue the image is blurry it hasn't been focused correctly either!
So many factors come into play for a sharp image, it's hard for a newbie to discern misfocus from camera movement, from subject movement, from ... etc.


I think Kevin hit the issue on the head here:


...... Thus it presents a rather unique scenario of being able to utilize the AF sensors for PDAF, and the image sensor for CDAF. With typical DSLRs, you either get 1 or the other :)

.....

Perhaps the camera is simply going "Hey, here's a face I detected. Please align it with one of the AF points so I can use PDAF for more accurate focusing"

I reckon what could(or at least should) be happening is that while the subject is off the AF grid(ie. on the image sensor and hence CDAF) .. it'll be followed and focused on using CDAF(when in face detect mode).
BUT ... as soon as the face lines up with an area in the AF grid, the PDAF sensor takes over, so it may appear that the face detect isn't working properly.

So are you confirming that if a face is nowhere near the PDAF grid, that if face detect is used, the camera simply will not focus on that face. Nor will it follow the face(as an object).
Remember too, CDAF is not always obvious in it's operation. I'd determine the colour of the focus area as an indicator of whether it's focusing, focused or not.

On Nikon cameras(well the D800 that I know of), a detected face in a green square is in focus, and a face in a yellow square is not(in focus).
it's actually a bit annoying(if you choose to use face detect that is), in Nikon choosing a yellow colour for the af square not in focus
In normal CDAF is, a focus square not in focus is red.
So the inconsistency can cause confusion. The difference in the yellow and green is pretty obvious, but for folks that have even a partial colour blindness, the difference between a green and a yellow may not be distinct enough.

[kevin301]

I believe you're right Kevin - there's just no way to get anything to be focused unless they're on one of the AF Points.
I've tried the portrait modes and movie mode - gotta be in the square.

So it's unfortunate - though using PDAF which works fast, there are still many limits such as a small area of focus, and if anything ever happens outside of it, or for convenience, it doesn't cut it.

Well, I guess you just work with what you have, but the other disappointing factor is that it only focuses 1 face at a time, rather than multiple, even if there are multiple faces lined up onto an AF point... :S

Well, case closed.

I wouldn't be too concerned about the camera not focusing on multiple faces at 1 time, i.e not having multiple boxes lined up on the faces. The nature of focusing is that it is by distance (from the camera). Therefore if all your subjects are aligned more or less along the plane of focus, getting 1 of the subjects to be in focus should cause the rest to be in focus as well.
Of course I say this very casually, ignoring some factors such as use of fast lens (e.g. f/1.4) or focusing on a near subject, which results in a particularly thin depth of field.

[bitsnpieces]

True that.

Though we talked a lot about Face Detection in the post, at the end of the day, it was trying to understand the autofocusing function of the a65 - so I think I understand how it works a lot better now, the capabilities, expectations, and how I should use it.

Thank you everyone for your inputs. :)