Canon EOS R5 Unite and Discuss!

Leigh wrote in post #19435284
Deep Learning was originally developed for the 1DX series...

1DX Mark III, to be specific. And only to work in Live view, since it needs the more detailed information from the image sensor to work with.
The 1DX can track certain colors with its 100 kilopixel color sensitive light meter, and apply focusing as long as the subject stays at one of the 61 AF points.
The 1DX Mark II can track faces with its 360 kilopixel light meter. Still limited to the 61 AF points.
But the 1DX Mark III can do a part of the tricks you find in the R3, as long as you use Live view.

I wonder if case 3 might be useful for horse racing. Lone riders are no big deal, but a pack of horses shoulder to shoulder makes tracking a bit more difficult.

apersson850 wrote in post #19435351
1DX Mark III, to be specific. And only to work in Live view, since it needs the more detailed information from the image sensor to work with.
The 1DX can track certain colors with its 100 kilopixel color sensitive light meter, and apply focusing as long as the subject stays at one of the 61 AF points.
The 1DX Mark II can track faces with its 360 kilopixel light meter. Still limited to the 61 AF points.
But the 1DX Mark III can do a part of the tricks you find in the R3, as long as you use Live view.

I thought that the 1Dx3 used its high-res metering sensor for some AF aid in DSLR mode? Obviously, it can't pick out a small eye like the newer Rs do, but the metering sensor should be able to image larger things.

goalerjones wrote in post #19435361
I wonder if case 3 might be useful for horse racing. Lone riders are no big deal, but a pack of horses shoulder to shoulder makes tracking a bit more difficult.

Are you in Face/Eye detect? Do you want a specific horse and rider in focus? If in Face/Eye STS will play more of a role. I don’t see too many obstructions. Anything that winds up in front of your subject will like I have an eye.

Also I have found subject to detect is very sensitive. I was on people and forgot to put it back to animals. Shooting animals I thought my camera was broken.

When in Face/Eye, multiple targets in on the frame and I want a specific target I use what I call the force focus method. The system will pick the best eye, which may not be the subject I want. I will use one of by BBF buttons and force the camera to focus on the subject I want. In this situation I’d use AF/ON which is mapped for single point AF. Usually placing the AF point anywhere on the body would do. Release the BBF and the eye snaps in. For a bunch of horses and riders I’d place the AF point closer the head area of my intended subject. It’s a very fast and accurate technique.

I’d set STS to 0 because I want the system to stick with the eye I select. There is a warning about 0 but there is not a lot of erratic movement in this example.

Later I’ll post a video of that technique in action.

I do have the link on my laptop. I used Case 3 with my DSLR's when the situation required it. Face/Eye has changed everything and I can't remember when was the last time I was not in that mode. Since getting my R and R5 I have not used Case 3.

Via BBF I use either Single Point, Zone or Spot AF. Single point used most often and Spot is used the leased. Watch the Fro use Expansion AF to force the camera to focus on the subject he wants.

It s Sony vs Canon R3 shootout but its' the same thing. Minutes 2 to 6 and as we know Canon uses blue AF points.

https://www.youtube.co​m/watch?v=LGrEWOZYgDw

Since there is so much information about Case numbers I've been posting a summery here for several years. Since some have DSLR's I left 5 and 6. I used them all but primarily 2 for single point and 6 for zone and I did tweak the parameters to my taste.

Case 1. Factory base settings that will cover most shooting situations.

Case 2. TS is negative.I set it to -2. If the AF point/s go off your subject AF will hang on and not refocus instantly on something else and give you time to reacquire your original subject. This includes your subject being obstructed momentarily by something else like a tree or another bird.  

Case 3. TS is positive.I set it to +2. The opposite of Case 2. System will immediately refocuses on another subject when AF point/s fall on it. The intent is that you are already focused on something else first. For example good for tracking two runners in a race, you are shooting both and are switching from one to another.

Case 4. Accel/Decel is increased. System is more sensitive to objects moving to towards and away from you which includes abrupt stops/starts and increase/decrease in speed from a start point to a finish point.  

The next two are really more suited for the zone AF modes. I use them exclusively in Zone AF modes only. Mostly Case 6 for birding.

Case 5. AF switching is increased. The camera more sensitive to left, right, up and down movement.

Case 6. The Grand Poobah. Combines Cases 4 and 5. Both Accel/Decel and AF switching are increased thus front to back and left/right up/down erratic movement is covered.

I got my 100-500 first so I used it with my R and shot some drag races. The cars were accelerating towards me so I used Case 4. I got my R5 about 6 weeks later.

John Sheehy wrote in post #19435368
I thought that the 1Dx3 used its high-res metering sensor for some AF aid in DSLR mode?

That's already in 1DX Mark II, so I didn't repeat it for Mark III. I only listed the new part.

Several times on here I've sung the praises of Canon's cheapest lens, the RF 50 f1/8 - the nifty fifty. It was the lens on my camera when I spotted this spider in its web. You have to delve into the R5's menu to switch to manual focus. First time I've done that and I was yet again pleasantly surprised with the result. Of course, this lens has its limitations - quite a few, actually - but I love it, not least because it is so light and simple. Surprisingly sharp, too. Anyone else here use this lens?

Photo from Terrycanon's gallery.
Image hosted by forum (1180971)

Well this has been one of the best threads so far. Something occurred to me that I sort of knew. I never used Case 3 when Face/Eye AF and I can't ever see a need for it. Personally I think it is redundant.

If I'm shooting a play and the person I'm was tracking goes behind a short wall I would be using Case 2. Without getting into trouble with Levina and others I don't know why anyone would want to refocus on the wall. Case 2 would give some time for my subject to emerge again.

However as I stated earlier if my subject steps behind another person there is a face and eye for the camera to recognise. Now STS kicks in and reacts based on your settings.

I have asked before how would Auto know my intent. Do I want Case 2 or 3? I also said I thought Auto may be weighted towards Case 2. I'm actually now even more confident in that thought but I can't really back that.

So back to my original question. Technically it would be very easy for Canon to automatically incorporate Tracking Sensitivity with a negative value and Accel/Decel adjust to speed changes (which it already does in Auto). I've never trusted Canon 100% with early document releases. I have examples of how much better the 1DX II AF guide explained things compared to the 1DX and 7D2.

So I had that little guy on my shoulder saying what if Canon did incorporate those into Face/Eye. I am convinced that Canon did not but I have wondered about that since the day I got my R5.

Case 3 is for when you did not track something that you lost track of, because there was nothing to track to begin with. What you want to track suddenly emerges, and that's the first time you see it.
If you ever have any reason to do something like that, that I don't know. It's not frequent for me, if it occurs at all. I can usually reposition myself to see better in advance. But I can imagine situations where it would be appropriate.

apersson850 wrote in post #19435505
Case 3 is for when you did not track something that you lost track of, because there was nothing to track to begin with. What you want to track suddenly emerges, and that's the first time you see it.
If you ever have any reason to do something like that, that I don't know. It's not frequent for me, if it occurs at all. I can usually reposition myself to see better in advance. But I can imagine situations where it would be appropriate.

Yes.When in Case 3 the intent is for system to be focused on something first. For non Face/Eye modes I agree with your explanation. For Face/Eye AF I don't see any use for it because I'm working with people or animals.

I suppose if something interferes it can pick that up first and then find a face or eye later. In that situation I wouldn't depend on it. If a person interferes with my primary person I would expect it to pick up the new eye or ignore it depending how I have
STS set. If any of my desired outcomes did not happen I would use the force focus technique and make it happen.

Too many scenarios in Face/Eye AF to predict what people/animals may do. I know I don't want to refocus on a wall, fence, etc if my primary subject goes behind it so I'd be on Case 2. I'd set STS so it does what I want when person or animal interferes. If that failed I'd take over with a BBF and force focus. It's pretty much the way I have been shooting.

This fellow demonstrates how STS picked up the second critter. I can't see how Case 3 would help here. If that critter went behind a tree Case 2 would ignore it. I wouldn't want the system to refocus on the tree.

Minute 10:22

https://www.youtube.co​m/watch?v=5IKCU_q7vfE

At around the time you indicate he's discussing the same as you. Tracking something that's lost occasionally.
Case 3 is intended to address the situation where there is nothing to track to begin with, but suddenly it appears. Either you did suspect it to show up at a certain place (like that hidden jumping motocross driver I wrote about before) or you don't even know where, but just hope something will happen. So you quickly aim the camera to the appearing target and press the button. No previous tracking of the subject, but you want it to be tracked as soon as possible, so the series of photos you probably will take in such a case become sharp as soon as possible.

Notably these cases were developed before all these deep learning object detection and tracking functions were implemented. Perhaps it's more redundant now, or only useful if you for some reason select to turn all these new features off, thus effectively bring an R3 back to 1DX Mark II level. That I can't say. I've only had time to take a couple of thousand action shots with my R3 so far, so I'm just scratching on the new feature's capabilities.

apersson850 wrote in post #19435631
At around the time you indicate he's discussing the same as you. Tracking something that's lost occasionally.
Case 3 is intended to address the situation where there is nothing to track to begin with, but suddenly it appears. Either you did suspect it to show up at a certain place (like that hidden jumping motocross driver I wrote about before) or you don't even know where, but just hope something will happen. So you quickly aim the camera to the appearing target and press the button. No previous tracking of the subject, but you want it to be tracked as soon as possible, so the series of photos you probably will take in such a case become sharp as soon as possible.

Notably these cases were developed before all these deep learning object detection and tracking functions were implemented. Perhaps it's more redundant now, or only useful if you for some reason select to turn all these new features off, thus effectively bring an R3 back to 1DX Mark II level. That I can't say. I've only had time to take a couple of thousand action shots with my R3 so far, so I'm just scratching on the new feature's capabilities.

Good summary. Until we get more information I’ll just keep plugging along with what I’m doing now. Currently I find the force focus method the most useful process and users came up with that one.

digital paradise wrote in post #19435706
Good summary. Until we get more information I’ll just keep plugging along with what I’m doing now. Currently I find the force focus method the most useful process and users came up with that one.

I forgot to say that I will pay more attention to the STS. Primarily between 0 and 1. I registered it to the green menus a few weeks ago but I’ll look into button mapping. We have had company all week so I haven’t been near my gear.

goalerjones wrote in post #19435361
I wonder if case 3 might be useful for horse racing. Lone riders are no big deal, but a pack of horses shoulder to shoulder makes tracking a bit more difficult.

Though the EOS 1DX3 is a DSLR, & the EOS R5 Mirrorless, the AF Case Parameter settings, and explanations are identical, and that even includes the EOS 7D2.

Both the EOS 1DX3 AF Guide: http://downloads.canon​.com …_Mark_III_AFGui​deBook.pdf

And the EOS 7D2 AF Guide: http://gdlp01.c-wss.com …-Setting_Guidebook_EN.p​df

offer extensive explanation, and illustrated sequences, as to use for Case-3, as they do with the other CASE scenarios.

A specific R5, or R3 AF Guide would not likely differ in any substantial way.

Beyond that, an in-person Course, or accompanied aid from a Canon engineer might be the only remedy for full comprehension? Good luck with that!!!