Home · Register · Join Upload & Sell

Moderated by: Fred Miranda
Username  

  New fredmiranda.com Mobile Site
  New Feature: SMS Notification alert
  New Feature: Buy & Sell Watchlist
  

FM Forums | Canon Forum | Join Upload & Sell

1
       2       3       end
  

Archive 2021 · Canon R5 - Tracking Box is Inconsistent with Results - Animal Eye AF

  
 
Jesse Evans
Offline
• • • •
Upload & Sell: On
p.1 #1 · p.1 #1 · Canon R5 - Tracking Box is Inconsistent with Results - Animal Eye AF


There have been a number of posts on this forum regarding people having issues with the Canon R5 autofocus while using tracking + servo AF.

After some recent observations I have come to the conclusion that the box in tracking autofocus is not indicative of which point the camera is using for autofocus.

Or, alternatively, (and unlikely in my opinion) perhaps the blue box is not indicative of ALL of the points the camera is using for autofocus and it is using additional points that are causing issues.

So here is what has been going on.

Over the last few months, attempting to use Animal Eye AF with my dog, I had continuously gone through the following cycle:

1. Happy while out shooting, the blue box is on my dogs eye, typically she is in motion so it's hard to verify that the focus is correct.
2. Get home, and notice many, but not all, end up being focused on her nose rather than her eye.

A lot of the time I would be shooting with a 70-200 f/2.8L IS so it would be hard for me to judge whether it was just random error / misfocus due to the fact that she runs very quickly, so making any educated guesses about what was going on would be difficult.

The other day, my dog had a spaz moment. She pulled a blanket off of our fireplace insert, when it caused stuff to fall she freaked out and sprinted in full panic mode through the room. She unfortunately decided to run under our recliner, which I was sitting on in reclined position, and split her head open by hitting the metal support arms.

After taking her to the vet to have the wound trimmed and cleaned, I went to take a photo with my RF 50mm 1.2L to document the situation.

Well, while trying to wait for the right expression, I noticed that the focus of the lens would drift to her nose, even while the blue box was still accurately placed over her eye.

I didn't want to post here until I could collect video evidence of this phenomenon, because I feel that this would go over about as well as a ton of bricks.

Today, I took my Atomos V, and hooked it up to the HDMI port on my camera to record the EVF.

Here is the video showing this bizarre behavior

https://youtu.be/PjpGCd6Zq40

I don't exactly have a very good explanation for why this is happening. But what I can tell you is that I am very disappointed that this behavior exists.

Just so we are clear, if I change the AF to be 1 point, or expanded point, the AF point does not drift.

Anyway, I just thought I would share my findings. I intend to contact Canon support to inform them of this issue, hopefully it will be resolved.



Mar 19, 2021 at 07:23 PM
TeamSpeed
Offline
• • • • •
Upload & Sell: Off
p.1 #2 · p.1 #2 · Canon R5 - Tracking Box is Inconsistent with Results - Animal Eye AF


This would mirror the issues we have with mirrored cameras. The AF box isn't truly indicative of the sensor position or size that was being used for AF, the AF box was more of a suggestion of what the system was using, but the actual AF sensor array being used was different (ie. parts lying outside the box) than what the box shows.


Mar 19, 2021 at 08:34 PM
Jesse Evans
Offline
• • • •
Upload & Sell: On
p.1 #3 · p.1 #3 · Canon R5 - Tracking Box is Inconsistent with Results - Animal Eye AF




TeamSpeed wrote:
This would mirror the issues we have with mirrored cameras. The AF box isn't truly indicative of the sensor position or size that was being used for AF, the AF box was more of a suggestion of what the system was using, but the actual AF sensor array being used was different (ie. parts lying outside the box) than what the box shows.


I donít think you are completely incorrect in your assessment, but in this case I donít think there is much of a parallel to be drawn between mirrorless and mirrored.

The phase detection points are on the sensor, and are the pixels themselves. There is no reason to end up in a situation in which the af box is in a place other than where the autofocus is being activated. It is also entirely unpredictable when it will choose to simply focus on the nose or genuinely focus where the box is.

It is also unlike the behavior in any Sony or Fuji camera that I have owned.

I also suspect the other thread here regarding the bizarre behavior at swimming meets is also related to this behavior in the camera.

Iím going to work through this with Canon, but this behavior is, for me, a deal breaker.



Mar 19, 2021 at 09:06 PM
TeamSpeed
Offline
• • • • •
Upload & Sell: Off
p.1 #4 · p.1 #4 · Canon R5 - Tracking Box is Inconsistent with Results - Animal Eye AF


The boxes are just graphical indicators, and not the exact boundaries of the sensors being used. The graphical indicator programming to the user is likely uncoupled from the actual AF system and its algorithms. This is not unlike what we have for mirrored bodies where a box was just an indicator with the actual AF system being physically different in size.


Mar 19, 2021 at 10:04 PM
Jesse Evans
Offline
• • • •
Upload & Sell: On
p.1 #5 · p.1 #5 · Canon R5 - Tracking Box is Inconsistent with Results - Animal Eye AF




TeamSpeed wrote:
The boxes are just graphical indicators, and not the exact boundaries of the sensors being used. The graphical indicator programming to the user is likely uncoupled from the actual AF system and its algorithms. This is not unlike what we have for mirrored bodies where a box was just an indicator with the actual AF system being physically different in size.


Can you explain more about the point youíre trying to make? It seems to me you are making a statement that has nothing to do with the situation at hand, and Iíd like to understand what it is that youíre intending to convey.

The autofocusing points in the region represented by the box are definitively capable of accurate focus in that region. I donít really understand what it has to do with the lack of specificity of the ospdaf sensor etchings in a dSLR. When one of those points is highlighted, that is the point being used for AF.

That is definitively not the case here.



Mar 19, 2021 at 10:24 PM
Andrew J
Offline
• • • • •
Upload & Sell: On
p.1 #6 · p.1 #6 · Canon R5 - Tracking Box is Inconsistent with Results - Animal Eye AF


Very nice test. The tracking is very sensitive to exposure and the histogram shows no data in the brightest level. That level has half of all the data the camera can collect.


Mar 20, 2021 at 02:30 AM
TeamSpeed
Offline
• • • • •
Upload & Sell: Off
p.1 #7 · p.1 #7 · Canon R5 - Tracking Box is Inconsistent with Results - Animal Eye AF


Jesse Evans wrote:
Can you explain more about the point youíre trying to make? It seems to me you are making a statement that has nothing to do with the situation at hand, and Iíd like to understand what it is that youíre intending to convey.

The autofocusing points in the region represented by the box are definitively capable of accurate focus in that region. I donít really understand what it has to do with the lack of specificity of the ospdaf sensor etchings in a dSLR. When one of those points is highlighted, that is the point being used for AF.
...Show more


After reading and re-reading your opening post, when you say the camera seems to be using other AF points or doing something contrary to how you have your AF mode set up, then, yes I am not on the same page as your opening post (at leas part of it) and I apologize.

If that is happening, I have no answer for why Canon's system would do that, where it uses DPAF sensors so far off the blue box boundaries. My post above is that there might be a fuzzy zone around the blue box boundary, but not that much. Especially with Eye AF where it is supposed to keep boxes on the eyes and move that for you. Perhaps the nostrils on the dog is being interpreted as eyes from time to time?

I was trying to describe how the blue box is just a graphic that shows you roughly the area of the sensor and its DPAF sensors being used for AF and how those could fall outside the boundaries of the blue box. You stated that in the early part of your post (which is what I fixated upon), but later you talk more about how the system just isn't super consistent with its behavior of Eye AF and keeping focus on the eyes vs other things.


After some recent observations I have come to the conclusion that the box in tracking autofocus is not indicative of which point the camera is using for autofocus.

Or, alternatively, (and unlikely in my opinion) perhaps the blue box is not indicative of ALL of the points the camera is using for autofocus and it is using additional points that are causing issues.



Mar 20, 2021 at 04:52 AM
Bruce n Philly
Offline
• • •
Upload & Sell: Off
p.1 #8 · p.1 #8 · Canon R5 - Tracking Box is Inconsistent with Results - Animal Eye AF


Your camera is broken... send it back for repairs.

My R5 does not do that..... and I think it flaky every now and then even now, after learning it. When a blue box is on a subject, it is in focus. If it is not in focus, the box changes.... the camara finds something else... it doesn't stay on a point with the focus changing.... nope, your camera is broken. I never experienced this.

It is fun to speculate and learn how this stuff works under the covers, but unless you work for Canon in their R&D center.... it is all speculation. Fun, but speculation.

Peace
Bruce in Philly

Edited on Mar 20, 2021 at 09:40 AM · View previous versions



Mar 20, 2021 at 07:57 AM
armd
Offline
• • • •
Upload & Sell: On
p.1 #9 · p.1 #9 · Canon R5 - Tracking Box is Inconsistent with Results - Animal Eye AF


Jesse Evans wrote:
Can you explain more about the point youíre trying to make? It seems to me you are making a statement that has nothing to do with the situation at hand, and Iíd like to understand what it is that youíre intending to convey.

The autofocusing points in the region represented by the box are definitively capable of accurate focus in that region. I donít really understand what it has to do with the lack of specificity of the ospdaf sensor etchings in a dSLR. When one of those points is highlighted, that is the point being used for AF.
...Show more

Precisely. What is the OP's point? He states that, "There have been a number of posts on this forum regarding people having issues with the Canon R5 autofocus while using tracking + servo AF." Yes, technically there have been posts from people regarding AF though the number has been modest and by-in-large come from those who are initially using the camera. I don't hear these complaints from longer term users and if there were problems, the number of cameras being returned would eclipse sales.

Quite frankly, I use this camera for wildlife and sports; two of the most demanding AF situations. While the R5 is not perfect, it is the most consistently accurate AF'ing body I've ever owned from Canon. Period. Sure, my 1d bodies acquired af more quickly though because they were DSLR's they exhibited more drift and inaccuracy. No, they didn't suffer from the limitations of on chip AF, namely an inability to acquire af of objects completely oof when the lens was either racked near or far.

The dude speaks of a 70-200 f/2.8 and 50 f/1.2 as being his determining lenses and posts a video shot in extremely low light where the af appears fine initially until he moves in closer to perhaps beyond the MFD of the lens. When he backs away, the camera seems to recapture focus. So, what's the point? His combination doesn't work in near darkness when he moves within the MFD?

I'm not trying to be flippant, mean, or a fanboy but these claims are simply unreasonable as presented.



Mar 20, 2021 at 08:32 AM
TeamSpeed
Offline
• • • • •
Upload & Sell: Off
p.1 #10 · p.1 #10 · Canon R5 - Tracking Box is Inconsistent with Results - Animal Eye AF


Bruce n Philly wrote:
Your camera is broken... send it back for repairs.

My R5 does not do that..... and I think it flaky every now and then even now, after learning it. When a blue box is on a subject, it is in focus. If it is not in focus, the box changes.... the camara finds something else... it doesn't stay on a point with the focus changing.... nope, your camera is broken. I never experienced this.

It is fun to speculate and learn how this stuff works under the covers, but unless you work for Canon in their R&D center.... it is
...Show more

I doubt you have your settings and modes set up the same way, nor have you done this exact same test, so how can you really know that your camera won't work the same way? I know for certain I have had cases where the image I captured didn't have in focus what the blue box covered. The cameras aren't broken, they just aren't perfect algorithimically and they all run the same code, therefore they would all work identically or nearly so baring hardware differences/tolerances.

=====


The video at the 40s mark is a bit odd, but I suspect all the R5/R6 bodies would behave the same way. Also there is no issue of MFD in that part of the video, because the nose was closer and in focus, the eyes were not which is where the box was laying.



Mar 20, 2021 at 09:54 AM
dwweiche
Offline
• • •
Upload & Sell: Off
p.1 #11 · p.1 #11 · Canon R5 - Tracking Box is Inconsistent with Results - Animal Eye AF


Always interesting when we all watch the same video yet see different things.

There are obvious sections of the video where the camera is focusing closer than the displayed AF tracking point, so minimum focus distance is not a variable in this example.

It is a lower light level condition, but isn't AF rated down to like -6 EV? I remember seeing the early release videos showing the system finding and tracking the eye of a lion in near darkness.

Maybe eye detect is also looking for some contrast in the eye itself (i.e. pupil) and that's not present in this video, but that's just speculation, and certainly doesn't explain why the camera is clearly showing the AF zone tracking a completely out of focus area. It is obviously showing the eye detect algorithm is finding the eye.

I have no idea if this is a particular defect of this camera or if it's just a limitation of the system, but regardless, it is not what I would expect to see. Asking Canon makes sense. Good luck.



Mar 20, 2021 at 10:23 AM
armd
Offline
• • • •
Upload & Sell: On
p.1 #12 · p.1 #12 · Canon R5 - Tracking Box is Inconsistent with Results - Animal Eye AF


TeamSpeed wrote:
I doubt you have your settings and modes set up the same way, nor have you done this exact same test, so how can you really know that your camera won't work the same way? I know for certain I have had cases where the image I captured didn't have in focus what the blue box covered. The cameras aren't broken, they just aren't perfect algorithimically and they all run the same code, therefore they would all work identically or nearly so baring hardware differences/tolerances.

=====

The video at the 40s mark is a bit odd, but I suspect all the
...Show more

I rewatched the video again and you are correct that it doesnít appear that heís within mfd. however, heís shooting a dark eye in low light at f/1.2. Is the af perfect? No. Does he have focus assist which is the IR setting used to maximize af in these types of settings on or off? Again, I think this ďtestĒ hardly demonstrates the utility or capability of the camera. In circumstances like this, the user might be better served by adding artificial light to improve the focusing capabilities, using single point or mf.



Mar 20, 2021 at 10:26 AM
TeamSpeed
Offline
• • • • •
Upload & Sell: Off
p.1 #13 · p.1 #13 · Canon R5 - Tracking Box is Inconsistent with Results - Animal Eye AF


I can see the low light issue for sure, it would be good to see this test in good lighting.


Mar 20, 2021 at 11:43 AM
bipock
Offline
• • • •
Upload & Sell: On
p.1 #14 · p.1 #14 · Canon R5 - Tracking Box is Inconsistent with Results - Animal Eye AF


I agree with TS. It's been noted over the years with DSLRs that the AF box isn't necessarily 100% accurate in that the bulk of the points being use are in the box but not necessarily all of them.

I believe this follows with the R5 and have noted it in my use. Just because the focus box is on something doesn't mean that something is perfectly in focus, it may be slightly off. I notice this most often when I have a focus point on a swim cap but one tiny corner hangs over the background water. The focus system will pick the water. Not sure why as I would think this covered in the whole "learned AF" system, but I'm not 100% up-to-date on that one.



Mar 20, 2021 at 12:34 PM
Jesse Evans
Offline
• • • •
Upload & Sell: On
p.1 #15 · p.1 #15 · Canon R5 - Tracking Box is Inconsistent with Results - Animal Eye AF




armd wrote:
I rewatched the video again and you are correct that it doesnít appear that heís within mfd. however, heís shooting a dark eye in low light at f/1.2. Is the af perfect? No. Does he have focus assist which is the IR setting used to maximize af in these types of settings on or off? Again, I think this ďtestĒ hardly demonstrates the utility or capability of the camera. In circumstances like this, the user might be better served by adding artificial light to improve the focusing capabilities, using single point or mf.


Do you really think it is that low light?

Itís ISO100 at 1/100 f/1.2. That is 7.17 EV. Accounting for a 2/3 stop under exposure you still have plenty of light. Even if I had the illumination lamp on it would do nothing.

I can see how the recording of the EVF looks dark, so Iíll try and remedy that in future recordings.

I will be doing additional tests in outdoor lighting with different lenses as well.

Please donít assume I am trying to attack Canon or diminish the R5. My goal is collect conclusive evidence of the issue so that it can be fixed by Canon. It is very obviously a software issue.



Mar 20, 2021 at 12:41 PM
Jesse Evans
Offline
• • • •
Upload & Sell: On
p.1 #16 · p.1 #16 · Canon R5 - Tracking Box is Inconsistent with Results - Animal Eye AF




bipock wrote:
I agree with TS. It's been noted over the years with DSLRs that the AF box isn't necessarily 100% accurate in that the bulk of the points being use are in the box but not necessarily all of them.

I believe this follows with the R5 and have noted it in my use. Just because the focus box is on something doesn't mean that something is perfectly in focus, it may be slightly off. I notice this most often when I have a focus point on a swim cap but one tiny corner hangs over the background water. The focus system
...Show more

Have you noticed this happening when you are in single point AF mode? Because I have not been able to reproduce this at all using single point, expanded, or spot AF. It only happens in tracking AF mode.



Mar 20, 2021 at 12:43 PM
lighthound
Offline
• • • • • •
Upload & Sell: Off
p.1 #17 · p.1 #17 · Canon R5 - Tracking Box is Inconsistent with Results - Animal Eye AF


Looks like user error (settings) or a defective camera or a lens that is need of a FW update. Hardly anything that is widespread or indicative of a design flaw. I've had the R5 since day one and have never seen this behavior unless I simply screwed up. The AEAF blue box will sometimes jump off target of course just like any other brand does but when it does, the focus shifts accordingly.
I've noticed this behavior when the AF system gets (as I call it) "bored" and looks for something else when there is no movement by the subject or camera.

Just like any AF system, you have to feed them enough light. I have learned that when using AEAF, this is even more important in order to get consistent tracking.

I agree with what TS is talking about as I recall reading about this when doing some AFMA work on a DSLR. The artical said to not trust the visual location of the AF box as this can vary from one camera/lens to the next as it's simply a reference/visual aid for the user. However, I don't think the gross shift in focus as seen in that one section of video falls into this category because the focus & the box was on the eye and then the focus shifted while the box stayed on the eye. That is a bit bizarre behavior.



Mar 20, 2021 at 01:52 PM
bipock
Offline
• • • •
Upload & Sell: On
p.1 #18 · p.1 #18 · Canon R5 - Tracking Box is Inconsistent with Results - Animal Eye AF


Jesse Evans wrote:
Have you noticed this happening when you are in single point AF mode? Because I have not been able to reproduce this at all using single point, expanded, or spot AF. It only happens in tracking AF mode.


I've noticed in single point and spot. I had to quit using single point for my application is was so sketchy.



Mar 20, 2021 at 03:08 PM
armd
Offline
• • • •
Upload & Sell: On
p.1 #19 · p.1 #19 · Canon R5 - Tracking Box is Inconsistent with Results - Animal Eye AF


Jesse Evans wrote:
Do you really think it is that low light?

Itís ISO100 at 1/100 f/1.2. That is 7.17 EV. Accounting for a 2/3 stop under exposure you still have plenty of light. Even if I had the illumination lamp on it would do nothing.

I can see how the recording of the EVF looks dark, so Iíll try and remedy that in future recordings.

I will be doing additional tests in outdoor lighting with different lenses as well.

Please donít assume I am trying to attack Canon or diminish the R5. My goal is collect conclusive evidence of the issue so that
...Show more

Personally, I couldn't care less if you attack Canon or their products. I watched your video again and read through the comments and still have the following thoughts.

1) It appears as though it was shot under low light (underexposed in video) though you seem to present otherwise. Shooting wide open is tough for any AF system. An ev of 7 can be challenging and comparatively in wildlife or sports one is typically working in higher ev's.
2) Dark, low contrast eye is still a factor and the R5 exhibits as does any camera with AED difficulty with these situations (including Sony; I've shot the a7riv and a9(ii) and they struggle as well in these circumstances).
3) I've used the R5 in thousands of different lighting conditions, levels, and subjects and it is the most accurate AF system of any Canon camera I've owned to date. Again, it's not perfect and is subject to the limitations of MILC.
4) You've not indicated whether you have the AF assist (IR) on or off. Repeat the experiment and let us know whether it makes a difference.
5) For those of us who work with the camera in the WL/Sports world, "pumping the af" is a frequent technique employed when using tracking and the desired af point is not captured or is lost. Alternatively, you may want to consider not using AED and tracking in the circumstance you demonstrated (single point, zone, etc.).
6) At the beginning of the video, the tracking system appears to be working fine. It follows the eye, jumps to a different spot when the dog moves etc. Quite frankly, I haven't seen the behavior which is exhibited about 40 seconds when you and the subject are moving, the point jumps around and the lens racks oof. As you continue, the af situation appear to degrade and get much worse. Again, rather than simply holding the button, when the af gets wonky, reacquire by pumping the button.
7) Settings do make a difference and it is important to adjust the Case type and sensitivities.
8) Is your camera defective? That's possible and there is no harm sending it in to Canon along with your video demonstrating the behavior. I still don't think this is a definitive test. Take the camera outside and have a person walk/run towards you, side to side, away, etc. Do the same with your pooch. It may give you a better indication.



Mar 20, 2021 at 03:35 PM
TeamSpeed
Offline
• • • • •
Upload & Sell: Off
p.1 #20 · p.1 #20 · Canon R5 - Tracking Box is Inconsistent with Results - Animal Eye AF


Yeah I pump the AF button habitually during a game, I don't even think about it. It is a habit formed from the DSLR world, but it isn't needed as much with mirrorless. This is evidenced by ES running 20fps tracking a player with nary a frame OOF.


Mar 20, 2021 at 05:13 PM
1
       2       3       end




FM Forums | Canon Forum | Join Upload & Sell

1
       2       3       end
    
 

You are not logged in. Login or Register

Username       Or Reset password



This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.