• Welcome to BirdForum, the internet's largest birding community with thousands of members from all over the world. The forums are dedicated to wild birds, birding, binoculars and equipment and all that goes with it.

    Please register for an account to take part in the discussions in the forum, post your pictures in the gallery and more.
ZEISS DTI thermal imaging cameras. For more discoveries at night, and during the day.

Why no 'Eye Directed' AF ? (1 Viewer)

Chosun Juan

Given to Fly
Australia - Aboriginal
Why has no-one made an AF system directed by our eyes? (ie. what we are looking at) since the subject of our focus (pun if you like :) is invariably where we would like the camera to focus too.

I realize there are a couple of issues involved:
* Our eyes can often dart around the scene - checking composition, or other incidental subjects in the scene (such as a prey animal being chased by predator etc) ,
* Or, the subject we would like to image may be moving too quickly and/or erratically for our gaze to stick rigidly to it - think swallows, hummingbirds, or dragonflies darting around etc
* Or our gaze at times may be a bit non-specific - focusing on a whole bird for example rather than a specific point such as its eye.
* The shutter interrupts our vision (and obscures instantaneous movements which may occur during the blackout period),
* Etc,

DSLR'S have dedicated phase detect sensors and elaborate AF and tracking using RGB sensors. Nikon's '3D Tracking' is currently king of the heap (especially in the D5). Canon's TTR system is displayed at its best in the IDX MKII. Mirrorless probably look to the Sony A9 as the most proficient AF system and they have been catching DSLR's rapidly. With phase detect on chip, they offer nearly the entire image (93%) as available for focus points. Olympus's OMD EMI MkII also gets an honorable mention.

The Sony has 'Eye AF' where it will lock onto and track a subject's eye, but this is not really what I am talking about. My old Minolta 700Si filum jobbie had 'Eye-Start AF' and as awesome as this was (beats a half depressed shutter for me - especially when each excessive pressure slip was a precious film frame potentially wasted), it falls short of what I am talking about too.

Some of the AF algorithms in current cameras must be pretty naff and unsophisticated. Even my D7200 which I regard pretty highly in general can jump from a panned falcon to an escarpment 50m in the background if I happen to wobble off the subject. Other cameras are much worse, hunting in ways that defy the laws of physics ! ..... perhaps they are allowing for that once in a lifetime capture of a UFO that jumps hundreds of meters or a kilometre or more instantaneously! :cat:

With some automotive safety systems now detecting whether our eyes are on the road or if we have nodded off, it would seem the technology at least in principle exists for an 'Eye Directed' AF. Perhaps it would even work in concert with traditional systems and algorithms ....

I think it would be pretty cool - how about you?

Thoughts and discussion welcome ......



Chosun :gh:
 
Well, it's not a new idea. Pre-digitally, I used to have a Canon EOS 50E SLR which had eye-controlled AF. In practice I just didn't use the facility - it seemed an unnecessary complication. Unfortunately I can remember next to nothing of the details, but as far as I'm concerned, this is something best left in the dustbin of history |=(|

Don't think I really have any more to contribute to this discussion, though it will be interesting to see what others think.
 
Hi Chosun,

Why has no-one made an AF system directed by our eyes? (ie. what we are looking at) since the subject of our focus (pun if you like :) is invariably where we would like the camera to focus too.

While the idea is fascinating, I think there are a number of factors that reduce its practical value.

One is that the human eye can move quicker than the camera can re-focus. I don't believe I always keep my eye centered on the bird in the view finder, for example when I'm waiting for it to display some specific aspect of behaviour.

Another is the limited accuracy of eye movement detection. The autofocus often has a hard time of focusing on a bird instead of a random branch in the tree near the bird, and that won't get any easier if there's some uncertainty where the desired autofocus axis is pointed.

Like Jonno52, I've (briefly) used a Canon with the eye movement autofocus feature many years ago, and I wasn't particularly impressed.

I could imagine that a semi-automatic mode would be useful ... I push a button, and while I'm holding it, I look at the spot where I want to put the autofocus measuring field, then I release the button. My camera requires invocation of some menu item and operation of the cursor keys, which takes so long that I never manage to pull that off in an action situation.

However, I doubt continuous eye tracking would be as useful in practice as one might think at first. I don't believe it would work as well in camera control as it works with the "Eyeball Mk I" ;-)

Regards,

Henning
 
Thanks for the replies John and Henning.

What I had in mind would be rather more sophisticated than those earlier Canon examples you cite. I am not personally familiar with them, but I believe they allocated focus in up to 7 'zones' ? Given that the Sony A9 etc is using 693 focus points covering 93% of the frame, I would envisage something offering at least that focus resolution if not several times (late now and an early start, so I will crunch the numbers another day unless someone does it first).

I am not sure that what Henning suggested about the focus not being able to keep up with the human eye strictly holds true (maybe that older system). The rapid movement of the eye (saccade) has a peak angular velocity of 900°/s https://en.m.wikipedia.org/wiki/Saccade

Some of the Japanese NC machinery used in electronics manufacture, even designed back in the 1980's, goes so fast that I doubt I could keep up with with a focused view when it is going full tilt. Certainly, from what I have seen, birds such as Peregrines, Swifts, and (Hummingbirds) etc changing direction at full tilt are similarly hard to follow. I'd like to see fit her firm numbers on the situation, but I doubt focus speed would be a limiting factor .....

Upon doing a bit more research https://en.m.wikipedia.org/wiki/Eye_tracking it seems that imprecision in pinpointing the exact point of focus (given that you would use non-invasive optical eye movement sensing) may be more of a limitation as Henning mentioned .....
I'm sure this could be overcome by developing an application specific solution, and so I'm still of the belief that such a system could work in a superior fashion to the best available AF systems available now. Even if it worked in conjunction with traditional phase detect systems, surely it would eliminate much of the huge "whoopsies !" that we see at the moment where the subject is totally lost and the lens ends up hunting into blurred oblivion.

Further thoughts anyone ?



Chosun :gh:
 
Hi,

I am not sure that what Henning suggested about the focus not being able to keep up with the human eye strictly holds true (maybe that older system). The rapid movement of the eye (saccade) has a peak angular velocity of 900°/s https://en.m.wikipedia.org/wiki/Saccade

I was thinking more of the "accomodation" of the camera lens itself, which can be a bit slow compared to the speed of the human eye. In fact, my thought was that maybe the eye would be able to jump so quickly between different points in the picture that the camera would "hunt" between the different focus settings required for these two points.

Certainly, from what I have seen, birds such as Peregrines, Swifts, and (Hummingbirds) etc changing direction at full tilt are similarly hard to follow. I'd like to see fit her firm numbers on the situation, but I doubt focus speed would be a limiting factor .....

From photographing swifts in full flight, focus didn't seem to be much of a problem (using my Panasonic DMC FZ1000). The main reason many pictures were blurred was tracking accuracy. I didn't even use the camera viewfinder, but a reflex sight to ease initial target acquisition. The sharpest pictures systematically where those where the swift was right at the edge of the frame, as that's when I transited zero relative motion.

Even if it worked in conjunction with traditional phase detect systems, surely it would eliminate much of the huge "whoopsies !" that we see at the moment where the subject is totally lost and the lens ends up hunting into blurred oblivion.

Being a technology enthusiast, I'm confident that new technology will introduce new huge "whoopsies" we can't even think of yet! ;-)

Maybe a semi-automatic mode, such as a "zone lock" to tell the camera not to go for sudden large focus changes, would do the job. Some SLR tele lenses have a "close range lockout" switch, which is sort of a low-tech version of this.

Regards,

Henning
 
Hi,



I was thinking more of the "accomodation" of the camera lens itself, which can be a bit slow compared to the speed of the human eye. In fact, my thought was that maybe the eye would be able to jump so quickly between different points in the picture that the camera would "hunt" between the different focus settings required for these two points.



From photographing swifts in full flight, focus didn't seem to be much of a problem (using my Panasonic DMC FZ1000). The main reason many pictures were blurred was tracking accuracy. I didn't even use the camera viewfinder, but a reflex sight to ease initial target acquisition. The sharpest pictures systematically where those where the swift was right at the edge of the frame, as that's when I transited zero relative motion.



Being a technology enthusiast, I'm confident that new technology will introduce new huge "whoopsies" we can't even think of yet! ;-)

Maybe a semi-automatic mode, such as a "zone lock" to tell the camera not to go for sudden large focus changes, would do the job. Some SLR tele lenses have a "close range lockout" switch, which is sort of a low-tech version of this.

Regards,

Henning

With the recent Nikon "Eye-AF" (focusing on the eyes - not 'eye directed' as I'm suggesting) firmware upgrade, this YouTube test shows what is state of the art with Mirrorless AF. https://www.youtube.com/watch?v=gxqa_f6LsQI

I have to say it's massively disappointing with some huge "whoopsies". I wouldn't expect my f2.8 lens to go on a flight of fancy and lurch off to something off centre waaaaaaay in the background. That just seems like a basic algorithmic ("zone lock" - I would have thought that this would be basic and remain bounded by the laws of physics .... apparently not - these systems seem to allow for picking up higher dimensional objects popping in and out of 3-D reality ! ) failure. How the heck would it go with the f5.6 and f6.3 lenses a lot of us use ? Even the king of the AF heap Sony has a frustrating time when changing eyes to focus on (the Nikon "ping-ponging" would drive me mental !)

I still say that a true eye directed AF with some algorithmic backup would offer a far superior and instantaneous system :cat:




Chosun :gh:
 
Last edited:
Canon introduced eye tracking AF (it focuses where the photographer's eye is looking through the viewfinder) in 1992 with the A2E. It stayed through the ELAN 7. I have a 7s with this feature, and it works pretty well.

Remember that film is much more forgiving for slight focus errors because the film plane is not flat. A digital system needs much better focus because the sensor is entirely planar compared to the thickness of film emulsion. So I am not sure how well you could get good sharp focus on digital. But digital might be able to combine the general are you are looking at with distance information and pick the most likely target in that area.

Marc
 
Eye Directed AF to appear on newly announced Canon R3 !
(must check mailbox for royalty cheque 😉)



Chosun ✊
 
Eye Directed AF Canon R3

The Canon R3 looks like a flagship by any other name. Initial impressions of the Eye Directed AF seem positive. There is also conventional AF selection, and it has the very useful toggle selector like the IDX III which I would love to see on all cameras.


I wonder if the Eye Directed AF will ever filter down to the R5 ? I could see myself tempted over to the Canon fold in that case ......

It will be interesting to see where the Nikon Z9 positions itself now - it seems like it would have been awesome 3 years ago, but will have it's work cut out for it now. Probably too late in the development cycle to include Eye Directed AF now. Canon is certainly lifting the bar .......


Chosun 🙅‍♀️
 
Canon R3 Eye-Control AF in practice ....


The fact that it combines with eye AF and let's you also set a focus zone, is really very enticing 👍


Chosun 👧
 
Warning! This thread is more than 3 years ago old.
It's likely that no further discussion is required, in which case we recommend starting a new thread. If however you feel your response is required you can still do so.

Users who are viewing this thread

Back
Top