CAF vs CAF+TR with Subject Detection with the OM-1.

I decided to try to see if there was difference was between CAF and CAF+TR when using Subject detection.

I took one of my swallow photos, made the bird smaller and placed it in a layer in Photoshop. I increased size the sky area as the background. This allowed me to move the bird in the sky to compare CAF and CAF+ TR to see how well each maintained focus on the bird as I moved the camera and bird around the frame.

CAF keeps the large rectangle around the target and attempts to keep the focus on the bird’s eye/head. CAF+TR drops the large rectangle with half press but still attempts keep the focus on the targets head.
The camera defaults to CAF+TR without Subject detection once focus is initiated.

This is easy to demonstrate when using Subject detection. If the camera is first focused on something other than a bird and the bird is moved to the center of the frame CAF+TR will not recognize or focus on the bird. With CAF mode, when the camera is initially focused on something other than a bird, it will immediately refocus on the bird when the bird is moved within the focus box.

For all different size focus areas, the outside focus area marks in the EVF accurately mark the maximum area scanned for the subject. As long as any part of the target is at the edge of the focus area, the camera will focus on the target, even if the rest of the target is outside the focus box.

If the camera or bird is moved, so that no part of the bird is in the focus area, CAF will now focus on the background. If the focus is set for Center Priority, the camera will focus closer to the middle of the frame. Without Center Priority, CAF will focus on the most detail area in the focus area.

CAF mode focus will jump to the background/foreground if you fail to keep the target in the focus area, but the camera should regain Subject identification focus once the target is within the focus box. If you keep the target within the focus box center, then Center Priority will decrease the likelihood of the focus jumping to the background when the camera fails to recognize the target. Larger focus areas make it less likely that you will fail to keep the target in focus, but make it more likely that the focus jumps to another target in the larger focus area.

Initial focus for both CAF and CAF+TR is constrained by the focus box area. However, once focus is acquired, CAF+TR will continue to keep the target in focus anywhere within the frame. Center Priority has no noticeable effect on CAF+TR.

CAF+TR will not focus on anything when the target is no longer in the frame, it will try to focus and then just stop trying to focus the target moved back in the frame.

CAF+TR does have another advantage (in addition to focusing outside the focus area once focus is obtained) in that it tends to just stop trying to focus when it loses the target, so it may be more likely to quickly regain focus with momentary failure to keep the target in the frame. However, since it does not appear to use Subject detection to maintain focus, it will be more likely to lose focus on a bird whose appearance keeps changing as the bird flies.



View attachment 90a2449fbee846398fad57e6332ce7ea.jpg



--
drj3

Will this be on the midterm?

Wonderfully complicated exploration. Thanks for the writeup.

Cheers,

Rick

Skeeterbytes said:

Will this be on the midterm?

Wonderfully complicated exploration. Thanks for the writeup.

Cheers,

Rick

Click to expand...

Too many years of college/university teaching statistics/research design to give a simple explanation.

drj3 said:

Skeeterbytes said:

Will this be on the midterm?

Wonderfully complicated exploration. Thanks for the writeup.

Cheers,

Rick

Click to expand...

Too many years of college/university teaching statistics/research design to give a simple explanation.

Click to expand...

I must admit I had difficulty following your writeup. Not sure how effective your test replicates a real live shooting scenario.

Could you please provide a a concise summary of what you concluded? What worked best?

Thanks.

--

Denjw said:

drj3 said:

Skeeterbytes said:

Will this be on the midterm?

Wonderfully complicated exploration. Thanks for the writeup.

Cheers,

Rick

Click to expand...

Too many years of college/university teaching statistics/research design to give a simple explanation.

Click to expand...

I must admit I had difficulty following your writeup. Not sure how effective your test replicates a real live shooting scenario.

Could you please provide a a concise summary of what you concluded? What worked best?

Thanks.

Click to expand...

The answer depends on the skill of the photographer.

Mirrorlesscomparisons used the large focus area with subject detection.

Always use CAF unless you have a single bird with the sky as a background where CAF+TR will probably work, but CAF with All focus points with Center Priority would probably still be better since CAF+TR doesn't use subject detection.

If you can keep the target in the center of the frame, smaller focus areas with Center Priority will give the highest percentage of images with good focus and this will prevent the focus from jumping to another bird. However, you must keep at least part of the target in the focus area, or you will completely lose focus on the target, which is the disadvantage of a small focus area.

As you increase the focus area, it becomes easier to maintain subject detection, but increases the probability of the camera focus jumping to the background or another bird in the background.

For very small, fast flying birds like swallows, using all focus points will probably increase the success rate for most of us since the bird can quickly change its location in the frame.

For small birds in trees/shrubs, I will use the single focus point with a front button controlling subject detection. This way I can quickly turn it off with a button tap when it results in a poorly focused target.

I don’t have an X or an OM-1, but it seems that the OM1’s CAF and CAF-Tr work the same as they do in the Mark II and Mark III and the only real difference is that Bird ID can find a bird so long as it’s within or touching the AF Target area

Peter

Messier Object said:

I don’t have an X or an OM-1, but it seems that the OM1’s CAF and CAF-Tr work the same as they do in the Mark II and Mark III and the only real difference is that Bird ID can find a bird so long as it’s within or touching the AF Target area

Peter

Click to expand...

It is the same in terms of CAF and CAF+TR when the camera does not detect a subject (or subject detection is turned off). When it does detect a subject, then that is what determines where the camera focuses as long as it touches the AF target area.

The subject detection is very good as long as the head/eye of the subject is not blocked. It can focus poorly if there are branches/twigs in front of the eye. I have a button assigned to subject detection so I can turn it off when necessary.

Subject detection is very good for the very active stationary bird where you have difficulty keeping the focus point on the eye in addition to its use for the flying/running targets.

drj3 said:

Messier Object said:

I don’t have an X or an OM-1, but it seems that the OM1’s CAF and CAF-Tr work the same as they do in the Mark II and Mark III and the only real difference is that Bird ID can find a bird so long as it’s within or touching the AF Target area

Peter

Click to expand...

It is the same in terms of CAF and CAF+TR when the camera does not detect a subject (or subject detection is turned off). When it does detect a subject, then that is what determines where the camera focuses as long as it touches the AF target area.

The subject detection is very good as long as the head/eye of the subject is not blocked. It can focus poorly if there are branches/twigs in front of the eye. I have a button assigned to subject detection so I can turn it off when necessary.

Click to expand...

+

drj3 said:

Subject detection is very good for the very active stationary bird where you have difficulty keeping the focus point on the eye in addition to its use for the flying/running targets.

Click to expand...

I’ve lost most of my interest in BIF and besides, my Canon DSLR is more than good enough for that.

But you’re right about the active ‘stationary’ little birds - shifting AF points left and right as the bird switches position on a branch can be difficult. Bird ID would make composition easier

Still not enough advantage for me to be willing to spend AU$4500 (incl. batteries, charger and HLD-10).

Your test reports are always interesting and thought provoking- thanks

Peter

Whoa!Can I get the Cliff Notes?

offtheback said:

Whoa!Can I get the Cliff Notes?

Click to expand...

+1

No, I always required my students to reach their own conclusions based on the data.

drj3 said:

Always use CAF unless you have a single bird with the sky as a background where CAF+TR will probably work, but CAF with All focus points with Center Priority would probably still be better since CAF+TR doesn't use subject detection

Click to expand...

I thought that the whole point of AF Tracking (regardless of which camera system one uses) would be to track one particular object amongst several objects (e.g. following one particular bird in the sky amongst a flock, or one bird as it temporarily passes behind trees or is obscured momentarily).

In short, I thought tracking on ANY camera was for following erratic objects that become temporarily obscured or leave the frame temporarily.

Isn't that what tracking is supposed to do? (Also thought it would try to reacquire the subject focus if the subject left the field of view / frame momentarily and then came back in to the FOV / frame.)

Either I am clueless as to the purpose of tracking or the OM-1 tracking is particularly unusual... or both.

--
What Middle School Is Really Like:

Off The Mark said:

drj3 said:

Always use CAF unless you have a single bird with the sky as a background where CAF+TR will probably work, but CAF with All focus points with Center Priority would probably still be better since CAF+TR doesn't use subject detection

Click to expand...

I thought that the whole point of AF Tracking (regardless of which camera system one uses) would be to track one particular object amongst several objects (e.g. following one particular bird in the sky amongst a flock, or one bird as it temporarily passes behind trees or is obscured momentarily).

In short, I thought tracking on ANY camera was for following erratic objects that become temporarily obscured or leave the frame temporarily.

Isn't that what tracking is supposed to do? (Also thought it would try to reacquire the subject focus if the subject left the field of view / frame momentarily and then came back in to the FOV / frame.)

Either I am clueless as to the purpose of tracking or the OM-1 tracking is particularly unusual... or both.

Click to expand...

OMD has stated that there were basically no changes in CAF+TR over the E-M1s. It was an algorithm that attempted to recognize the target and follow it while the camera predicted focus based on actual focus plus focus of previously exposed frames.

The weakness was always that targets change size and appearance as they move. It never worked very well for anything other than focus and recompose for stationary targets that did not change appearance.

The OM-1 has chosen a completely different approach. It can now focus extremely quickly, so OMD indicates that it no longer uses prior images in focus prediction. The tracking is now appears to be based on recognition of target type and focusing on that target type within the selected focus area.

You can change the "stickiness" of that with the C=AF sensitivity, whether there is a Center Priority and the focus area size. Users will need to report more about its performance with multiple birds/animals in the frame and how well it works in different situations.

That is an exquisite sample photo -- I know how fast those birds are and how hard to keep in frame, I've never succeeded at it myself. But the JPG is 87MP -- what did you take it with?

Matt Waller said:

That is an exquisite sample photo -- I know how fast those birds are and how hard to keep in frame, I've never succeeded at it myself. But the JPG is 87MP -- what did you take it with?

Click to expand...

Actually it is the attached uncropped image taken with the E-M1.3+MC20+300mm f4.

To make a test image, I reduced the size of the bird in separate layer of the E-M1.3 image and made the sky much larger. I also rotated the bird to see how that affected recognition.

I needed something with a simple bird image that I could move around the sky and check the location of subject identification limits.



--
drj3

drj3 said:

OMD has stated that there were basically no changes in CAF+TR over the E-M1s. It was an algorithm that attempted to recognize the target and follow it while the camera predicted focus based on actual focus plus focus of previously exposed frames.

The weakness was always that targets change size and appearance as they move. It never worked very well for anything other than focus and recompose for stationary targets that did not change appearance.

The OM-1 has chosen a completely different approach. It can now focus extremely quickly, so OMD indicates that it no longer uses prior images in focus prediction. The tracking is now appears to be based on recognition of target type and focusing on that target type within the selected focus area.

You can change the "stickiness" of that with the C=AF sensitivity, whether there is a Center Priority and the focus area size. Users will need to report more about its performance with multiple birds/animals in the frame and how well it works in different situations.

Click to expand...

Thanks for the explanation.

Still a great capture -- I love how there's a catchlight in the swallow's eye!

Matt Waller said:

Still a great capture -- I love how there's a catchlight in the swallow's eye!

Click to expand...

Thanks.

I hope there are more swallows this year, than last. I want to see if there is improvement over the E-M1.3 in success rate.