Patent: Canon shows off some information about the eye-controlled AF point selection

[Canon Rumors Guy]

The Canon EOS R3 will come with eye-controlled autofocus, made most famous by the film camera, the EOS 3. There have likely been a few advancements in technology over the last few decades, but there are some similarities in how the system will be calibrated.
The USPTO patent US 2021/0124239
The first patent shows the menu system that will likely be similar to what we see in the Canon EOS R3. It looks like there will be a lot of customization options to improve the experience.

The USPTO patent US 2021/0125375
The second patent shows off how Canon plans to set up calibration for eye-AF on the Canon EOS R3. This calibration is similar to the way the Canon EOS-3 worked, you will follow a moving target around the viewfinder. Canon has likely worked out a lot...

Continue reading...

 

[neuroanatomist]

Never used an EOS 3, but really looking forward to trying eye-controlled AF in the R3.

 

[Mt Spokane Photography]

I have the 7e, calibration was simple but reliability was fussy for me.

 

[WoodyWindy]

I had the EOS 50/Elan IIe, but I wear heavy glasses, so I'm not sure how that affected the eye control. I don't remember it being all that compelling. Of course, it didn't have nearly as many points to choose from as the EOS 3...

 

[t.linn]

I loved everything about my EOS 3 except the eye-controlled focus. It never worked reliably for my deep-set eyes.

 

[AJ]

I still have my A7E. The eye-controlled AF worked okay, but in practice I still ended up pointing, focusing, and then recomposing.
The R3 will be a whole different ball of wax. The old technology had the eye drive AF to fixed AF points. I think the new technology will be radically different. AI will determine points of interest in the photo, and with your eyes you'll be able to flip between these points. I have no doubt that the new implementation will be vastly more useful.

 

[calfoto]

While I’ve never had one of Canon’s cameras that utilized Eye Controlled Focus I can image there would have been a lot of errors with the technology they used at the time.

That being said I’m quite hopeful that the 2021 version will be a vast improvement - considering how well the R5 can detect and track a subject’s eyes. With some sort of Lidar like tech ECF may finally have come of age (fingers crossed )

 

[thomste]

Is the R7 still coming this year? Haven't heard anything about this for a while...

 

[unfocused]

Is the R7 still coming this year? Haven't heard anything about this for a while...

"still coming?" Was it ever coming? Don't confuse posts on a rumors site with actual facts

 

[scottburgess]

I bought the Elan IIe when it arrived, and loved the eye control focus (I've always worn glasses, but I'm near-sighted and the correction need is only moderately strong). I've been hoping for a return of this feature, as I found it worked beautifully if I simply concentrated on eye position prior to a half-press of the shutter button. Huge boon for wildlife work, hit the subject's eye every time. And I never struggled to calibrate the system. I wonder if eye color/contrast affects that?

I still have my Elan IIe. Maybe I should put some Ilford Delta 400 in it...

 

[unfocused]

I wonder about this. Canon seems to be marketing the R3 as a sports camera, but with certain sports, my eyes are constantly shifting around trying to anticipate the action. I suppose it will largely depend on the response time of the autofocus and it may be more useful for some sports than others. For example, in basketball or soccer, where you are following whomever has the ball it might work fairly well. With baseball, where I tend to spend most of my time scanning the field to be ready for plays that can occur in lots of different spots, it might not be so useful. I commend Canon for giving it a go and will be curious to see how well it works, although by the time this comes to market and is available at a reasonable price, I might be retired.

 

[David - Sydney]

I wonder how it would tell between different users of the camera.... did the EOS3 have a method for this?

 

[becceric]

I had the EOS 50/Elan IIe, but I wear heavy glasses, so I'm not sure how that affected the eye control. I don't remember it being all that compelling. Of course, it didn't have nearly as many points to choose from as the EOS 3...

I had the Elan IIe and thick lenses also. Unfortunately the results were quite hit or miss. I really did like the idea of their eye control depth of field preview, even with intermittent success. I would love to have a more successful version. There are many times while photographing flowers in the woods where it would be handy.

 

[dpockett]

I wonder about this. Canon seems to be marketing the R3 as a sports camera, but with certain sports, my eyes are constantly shifting around trying to anticipate the action. I suppose it will largely depend on the response time of the autofocus and it may be more useful for some sports than others. For example, in basketball or soccer, where you are following whomever has the ball it might work fairly well. With baseball, where I tend to spend most of my time scanning the field to be ready for plays that can occur in lots of different spots, it might not be so useful. I commend Canon for giving it a go and will be curious to see how well it works, although by the time this comes to market and is available at a reasonable price, I might be retired.

Exactly what I have been thinking, unsure how practical it will be with team sports.

 

[tapanit]

I wonder how it would tell between different users of the camera.... did the EOS3 have a method for this?

EOS 3 could store three different sets of calibration data that could be used for different users, or for same user with and without glasses for example. Selection between them was manual.

 

[tapanit]

I wonder about this. Canon seems to be marketing the R3 as a sports camera, but with certain sports, my eyes are constantly shifting around trying to anticipate the action. I suppose it will largely depend on the response time of the autofocus and it may be more useful for some sports than others. For example, in basketball or soccer, where you are following whomever has the ball it might work fairly well. With baseball, where I tend to spend most of my time scanning the field to be ready for plays that can occur in lots of different spots, it might not be so useful. I commend Canon for giving it a go and will be curious to see how well it works, although by the time this comes to market and is available at a reasonable price, I might be retired.

With EOS 3, the eye controlled focus point selection only engages when you first press the shutter (or *) button, it does not constantly track your eye movements. In AI Servo AF mode, only the initial focus point is selected by the eye control, after that it's automatic. How the R3 does it, I obviously don't know.

 

[unfocused]

With EOS 3, the eye controlled focus point selection only engages when you first press the shutter (or *) button, it does not constantly track your eye movements. In AI Servo AF mode, only the initial focus point is selected by the eye control, after that it's automatic. How the R3 does it, I obviously don't know.

Yikes! I hope the R3 will be continuous or at least offer that option. Can't see it being of much use for sports if not, but we will see.

 

[privatebydesign]

Yikes! I hope the R3 will be continuous or at least offer that option. Can't see it being of much use for sports if not, but we will see.

The idea was it was initial acquisition, just like area AF with initial spot selected, you choose the starting point then it automatically tracks the subject that was under that AF point all the time it is in the frame. It will 100% work in continuous.

 

[highdesertmesa]

I hope Canon uses some creative eye-controlled AF warning messages: "Too much baggage for optimal performance. Eye-controlled AF will be disabled until you get some sleep and lay of the sauce."

 

[Joules]

I wonder how it would tell between different users of the camera.... did the EOS3 have a method for this?

The EOS 3 implementation doesn't matter for this new version.

If you go back to the previous patent on this technology that CR guy posted, you'll see that the camera performs an iris scan using a small sensor and infrared light source inside the EVF. That is used to recognize a user. Besides calibration, the camera could also learn a user's preferences with regards to subject selection according to the patent. So if you have two people who like macro, but one shoots insects and the other flowers, the subject recognition would prioritize an insect on a flower for one user, and the flower itself for the other user.

I think that aspect of a patent was more them looking ahead at what could be done rather than something we'll actually see in the R3. But the method for observing the eye and obtaining the user's identity and what they are looking it are probably very close between the patent and the actual implementation.