E-M1.2 CAF: Focus Prediction Appears to be Based on Previously Exposed Images

paul cool said:

Think for small birds in flight the canon Nikon are better more consistent but unless they are top end dslrs or big cannikon primes the results look appreciably as bad as mine ,good luck with D500 and Tamron 150-600mm g2 and have fun trying capture swallows in flight with it .I am going to be trying my omd em1.2 with 300mm f4 and dot sight trying to capture small birds in flight if successful I will let you know even if 25% keepers I will not be racking up my shutter count and having nearly as good success rate due to higher frame rates.

Click to expand...

olyham said:

Bob657 said:

I’m beginning to wonder if there’s some variation between cameras or firmware, as some find CAF-Tracking the worst option (including me) and some hate while others love CAF..... go figure!

Click to expand...

Go read dpreview review of the EM1.2 my camera 'misbehaves' exactly as theirs did. Also there is many reports of this misbehaviour.

My camera was a fairly early one as would the Dpreview one if your camera does not misbehave in the same way as reported by all the threads including this one and it is a later one then it could be that you have a later motherboard that does not have the problem which sort of makes sense as Olys standard fix for AF problems is to chuck in another motherboard.

I must admit I was very surprised when some people reported improvements to CAF when they put in FW 2.0. and mine seemed to get worse.

I would be interested to know if you camera is an early one or recent or any others that don't have issues with CAF on the EM1.2 as it may be time to report the problem to Oly

Click to expand...

CliffWales said:

Tony Rogers said:

CliffWales said:

Adrian Harris said:

Stejo said:

CliffWales said:

Amazing timing, I’ve just come back from a full day shooting BIF, 720 images in total and out of those 3 in focus, yes just 3 and the detail on those 3 is frankly poor

i was so frustrated and annoyed with the focus acquisition and accuracy today, I,ve ordered the D500 and tamron 150-600 G2, I know I’ve threatened to do this before but this time I’ve done it

enough is enough, I’m just wasting my time , energy and frankly my mind on this , the Olympus omd em1 ii and PL 100-400 Is just not capable period.consistency with birds in flight, some days reasonable other days appalling, my photographer friend today nailed all the critical moments, every single one with his DSLRs ( D500 and 300 PF), I got none

my olympus kit will be retained for perched birds only and general photography but any serious wildlife and especially BIF will now be the Nikon,s domain

wildlife camera system ? Olympus? Dream on, it’s currently night and day the difference between the D500 and EM1ii, maybe the next iteration but not this version.Ah, I feel better

Rant over

Click to expand...

What kind of birds are you shooting? I'm sorry to say, but 3 keepers out of 720 shots suggest user error. Either in shooting technique or camera settings. Probably both.

Even with swallows in terrible light, my keeper rate with the GH5 and 100-300ii has never dropped below 25-30%. With large birds my keeper rate is consistently over 80%.

The EM1ii with the PL100-400 should be doing even better.

Click to expand...

Having just spent half a day in a hide with Nikon and Fuji users, I can assure Cliff that my success rate for really good keepers far exceeded theirs.

I was using em1-mk2 and pl100-400, they were using Nikon D4 + 200-500 lens and the Fuji latest cam + 100?-400.

They didn't have the reach or the focus acquisition speed. My em1-mk2 rarely messed up. And the interesting thing about shooting alongside others is you hear and see the truth about how their systems are performing compared to your own.

Cliff's problem sounds like user error to me.

Click to expand...

with respect, I am certainly not the only one who has experienced poor CAF performance with the EM1ii and 100-400 or indeed the 300 pro and think the Nikon D500 ( or 7dii) is far far superior as a camera for BIF, its well documented in articles, reviews, blogs and of course forums, as for my own alleged user error -you may well be right and this may has some bearing on the failure rate however, I didn't seem to suffer the same with the D7200 and 200-500 and my BIF skills have not changed that much to my knowledge LOL

an example is this one I found at random on DP review just now

https://www.dpreview.com/forums/thread/4101432

the whole of the first page is relevant especially poignant are the comments made by wildlifr • Senior Member • Posts: 1,611

in respect of your experience with the fuji and nikon users you mention and your superior keeper rate, all I can say is well done, clearly, kudos to you although I suspect at least with the D4 something is clearly amiss ,

can’t comment on the fuji as my only experience was with the XT1 and that was poor in CAF.

Click to expand...

I think that you are pretty much correct. E-M1 MkII is hard to get results with. Against a clear sky is the easiest but if the bird is against trees and reeds I have to deploy every trick I can - smallest possible focus group, set the limiter to try to eleiminate the background, have all the other settings just right, and track just about perfectly. Even doing that, I will probably fail. The main reason is that the E-M1 MkII doesn't favour the nearest subject. But even if I persuade it to focus on what I want, the accuracy is often poor. Very few of my best bif shots are against a cluttered background.

The E-M1 MkII was praised on release because it was a huge step up from the Mk I. Initially, I felt I finally had the tool to do the job. But, with experience, it is really a matter of degree. .i.e. how hard is it to use and how accurate are the AF results. And the answer is, it should be a lot better. The Mk I sometimes focused but was nearly always soft. The Mk II is ten times better but that still leaves it short.

I saw the "New WIldlife Camera" thread that you linked to but I decided not to comment because I didn't want somebody to spend all that money on an E-M1 MkII + 300mm based on my recommendation. I wouldn't recommend it to a friend so neither would I to a stranger. I don't have experience of dslrs but I have seen images, snatched in an instant, that I know I couldn't have got. E.g. a little bittern flying low over the water between reed beds.

I'm still hoping Olympus will release a C-AF update that favours the nearest subject rather than the most contrasty.

Click to expand...

appreciated Tony - your findings concur with my own, and probably articulated better as well, I certainly Hope Olympus can update the firmware to rival DSLR's as I much prefer to shoot with the Olympus but for now I'm awaiting receipt of the D500 + 150-600.Incidently how do you find using the 40-150 auto focus accuracy for BIf at close distance ( if you have tried it)

Click to expand...

Messier Object said:

Tony,

There are far too many similar reports for user error to be a major factor. And many who have disappointing results are experienced shooters who get better results with other cameras.

Click to expand...

Messier Object said:

I only use Olympus 4/3 lenses for shooting birds and while the 300mm F2.8 delivers superb image quality its AF performance is well below that of the 300mm F4 Pro. After much practice and experimenting with settings, and learning from others here , the best BIF keeper rate I get is about 20%. By most reports the 300mm Pro does a lot better. Yet I can pick up my Canon 7D Mk II and it will acquire, lock and Track a BIF and deliver 80-90% keeper rate with ease, and if the bird flies in front of or behind a tree the camera continues to track it. In clear sky I have on occasion shot a 30-frame BIF sequence with 28 or 29 frames sharp in focus.

I do not know what stops the E-M1 Mk II from equaling what the Canon can do, or the D500. Is it a limitation of the on-sensor PDAF ? Is it a matter of insufficient CPU power, ineffective software/algorithm, or some other limitation within the Olympus design team ?

Some here will say that the E-M1 II AF system can do anything the DSLRs can do. If that were true there’d be no complaints about its performance

Click to expand...

Messier Object said:

Peter

Click to expand...

Adrian Harris said:

CliffWales said:

Tony Rogers said:

CliffWales said:

Adrian Harris said:

Stejo said:

CliffWales said:

Amazing timing, I’ve just come back from a full day shooting BIF, 720 images in total and out of those 3 in focus, yes just 3 and the detail on those 3 is frankly poor

i was so frustrated and annoyed with the focus acquisition and accuracy today, I,ve ordered the D500 and tamron 150-600 G2, I know I’ve threatened to do this before but this time I’ve done it

enough is enough, I’m just wasting my time , energy and frankly my mind on this , the Olympus omd em1 ii and PL 100-400 Is just not capable period.consistency with birds in flight, some days reasonable other days appalling, my photographer friend today nailed all the critical moments, every single one with his DSLRs ( D500 and 300 PF), I got none

my olympus kit will be retained for perched birds only and general photography but any serious wildlife and especially BIF will now be the Nikon,s domain

wildlife camera system ? Olympus? Dream on, it’s currently night and day the difference between the D500 and EM1ii, maybe the next iteration but not this version.Ah, I feel better

Rant over

Click to expand...

What kind of birds are you shooting? I'm sorry to say, but 3 keepers out of 720 shots suggest user error. Either in shooting technique or camera settings. Probably both.

Even with swallows in terrible light, my keeper rate with the GH5 and 100-300ii has never dropped below 25-30%. With large birds my keeper rate is consistently over 80%.

The EM1ii with the PL100-400 should be doing even better.

Click to expand...

Having just spent half a day in a hide with Nikon and Fuji users, I can assure Cliff that my success rate for really good keepers far exceeded theirs.

I was using em1-mk2 and pl100-400, they were using Nikon D4 + 200-500 lens and the Fuji latest cam + 100?-400.

They didn't have the reach or the focus acquisition speed. My em1-mk2 rarely messed up. And the interesting thing about shooting alongside others is you hear and see the truth about how their systems are performing compared to your own.

Cliff's problem sounds like user error to me.

Click to expand...

with respect, I am certainly not the only one who has experienced poor CAF performance with the EM1ii and 100-400 or indeed the 300 pro and think the Nikon D500 ( or 7dii) is far far superior as a camera for BIF, its well documented in articles, reviews, blogs and of course forums, as for my own alleged user error -you may well be right and this may has some bearing on the failure rate however, I didn't seem to suffer the same with the D7200 and 200-500 and my BIF skills have not changed that much to my knowledge LOL

an example is this one I found at random on DP review just now

https://www.dpreview.com/forums/thread/4101432

the whole of the first page is relevant especially poignant are the comments made by wildlifr • Senior Member • Posts: 1,611

in respect of your experience with the fuji and nikon users you mention and your superior keeper rate, all I can say is well done, clearly, kudos to you although I suspect at least with the D4 something is clearly amiss ,

can’t comment on the fuji as my only experience was with the XT1 and that was poor in CAF.

Click to expand...

I think that you are pretty much correct. E-M1 MkII is hard to get results with. Against a clear sky is the easiest but if the bird is against trees and reeds I have to deploy every trick I can - smallest possible focus group, set the limiter to try to eleiminate the background, have all the other settings just right, and track just about perfectly. Even doing that, I will probably fail. The main reason is that the E-M1 MkII doesn't favour the nearest subject. But even if I persuade it to focus on what I want, the accuracy is often poor. Very few of my best bif shots are against a cluttered background.

The E-M1 MkII was praised on release because it was a huge step up from the Mk I. Initially, I felt I finally had the tool to do the job. But, with experience, it is really a matter of degree. .i.e. how hard is it to use and how accurate are the AF results. And the answer is, it should be a lot better. The Mk I sometimes focused but was nearly always soft. The Mk II is ten times better but that still leaves it short.

I saw the "New WIldlife Camera" thread that you linked to but I decided not to comment because I didn't want somebody to spend all that money on an E-M1 MkII + 300mm based on my recommendation. I wouldn't recommend it to a friend so neither would I to a stranger. I don't have experience of dslrs but I have seen images, snatched in an instant, that I know I couldn't have got. E.g. a little bittern flying low over the water between reed beds.

I'm still hoping Olympus will release a C-AF update that favours the nearest subject rather than the most contrasty.

Click to expand...

appreciated Tony - your findings concur with my own, and probably articulated better as well, I certainly Hope Olympus can update the firmware to rival DSLR's as I much prefer to shoot with the Olympus but for now I'm awaiting receipt of the D500 + 150-600.Incidently how do you find using the 40-150 auto focus accuracy for BIf at close distance ( if you have tried it)

Click to expand...

@ Tony and Cliff. I am so sorry that your m43 kit hasn't performed as you have hoped and is presenting difficulties and poor results.

What I find so strange is what is being deported is not my experience at all. Even with birds flying past reeds my em1-mk2 always Locks on instantly, however there are times when it will produce a few soft shots afterwards during a burst.

Click to expand...

Adrian Harris said:

I tend to only use C-AF in the default position for just about everything. But do notice that if very poor light the failure rate is more prominent.

I could no longer pick up a DSLR and long lens even if I wanted to, and certainly couldn't hand hold and shoot with it all day. But hope you soon find a solution to meet your needs.

Click to expand...

CliffWales said:

Tony Rogers said:

drj3 said:

Tony Rogers said:

First, thanks for doing the tests. We all need help trying to understand this!

Unless I missed it, you don't mention which AF points you used. Were you using a single point or more?

With you four sample images shown, I would be interested in knowing what the Exif attribute FocusDistance (and FocusStepCount) was for each of the four shots. My experience is that the camera wildly varies the FocusDistance between shots. Much more than you imply. It appears to oscillate around the correct focus.

Click to expand...

Have you checked this since firmware 2.0 with your 300mm? That update did improve CAF performance, though my experience would suggest that was mostly for use of CAF with stationary targets.

I will get back to you later with the requested information.

There were so many birds flying that the photographer fatigue from holding the camera out to use the Dot sight was the main problem.

I used all focus points and only birds with a sky as background.

I tracked only single birds (not difficult with the narrow field of view at 420mm). I did not want my ability to keep the focus point on the bird to add variability to the data.

I used the Dot Sight to better follow the birds for a longer period of time than would have been possible with the EVF. The Dot sight also enabled me to recover, if the bird moved quickly in a different direction out of the frame, so I could see what happened with lost focus (the same as initial focus).

Click to expand...

Thanks for the AF point info. That eleiminates losing the subject momentarily.

However, I have my concerns about how C-AF uses multiple points. It would be okay if the AF looked at all the points and picked the nearest one in each C-AF cycle and then used that for its predictions. However, I have a feeling it doesn't change points that often. It seems to stick with one/some for a while and then change. Sometimes the AF conformation rectangle disappears completely for a while. I can repeatable produce this by panning along a barbed wire fence at an angle to your line of view. If I pan too fast, the AF confirmation box is lost. Not quite sure what that means but the focus is poor when it does.

I look forward to the Exif data.

BTW My own take on the E-M1 MkII C-AF is that maybe it just isn't very accurate. Maybe the small, on-sensor pdaf points limit its resolution? On top of that, I suspect that the Olympus engineers have deliberately introduced a focus "wobble" around the approximate focus position so that at least some shots have a chance of being sharp, even though the measured AF location is rough. I have also never seen any evidence (before yours) that the C-AF uses any kind of prediction. I tend to half-press until I see the AF work and then press the shutter and often the first shot in the burst is the sharpest.

In short, I have never managed to convince myself that any of the settings available have an appreciable effect on the accuracy of the C-AF. I wish they did! My view has always been that the easiest subject to get accurate C-AF on is a stationary one. An algorithm that follows/predicts changes in distance can easily cope when that change in distance is zero. But using C-AF on a stationary subject yields a series of frames each of which has varying focus! I just tested this by pointing at my neighbour's roof apex with single AF point. Six shots and FocusDistance was 29.59m, 29.59m, 29.815m, 30.435m, 30.045m, 29.74m. So, a variation of 0.845m or getting on for 3 feet!

Click to expand...

absolutely, well thought out and articulated, this is exactly the issue, the camera focusses at random, and not always on the intended target, I put it down to focus shift and this can also happens on stationary subjects in burst mode, I've noticed a sharp image can be produced anywhere within the burst (or sometimes not at all, slightly out of focus )but unusually never all frames in perfect focus.

Click to expand...

drj3 said:

Distance from the camera for all images 17-25 for images 19-22 example in the initial post are given at the end of this post.

All images 17-25 had a green square indicating the camera confirmed focus in Live View. All images except #20 & #21 were in acceptable focus. Images #20 is somewhat out of focus (but the tail is close) and #21 was completely out of focus.

The camera estimate focus for image #21 is clearly wrong and the image is definitely not in focus. Something in images #20 & #21 must have caused the camera to predict a closer target than the actual bird. As can be seen in the image, the bird is actually accelerating away from its previous course and not coming closer to the camera. That can cause a serious problem for a bird that can covers between 2.25 and 4 feet in 1/15 second. However, the camera focused the next image 47.44 inches from #21 giving an acceptably focused image for image #22.

The bird images in this series only occupy .27MP to .81MP of the 20.2MP images, so there is not a lot of detail. I normally would not photograph a bird as small as a swallow from this distance other than for a camera focus evaluation.

The image post is repeated in the first image.

Focus distances in meters as reported by ExifTools are listed below and the graph is shown in the second image.

#17 – 22.275, #18 – 22.62, #19 – 22.795, #20 – 22.885, #21 – 22.705, # 22 – 23.91, #23 – 24.51, #24 – 24.72, #25 – 24.875.

The difference between successive images in inches is shown in the second graph (3rd image).







--
drj3

Click to expand...

Tony Rogers said:

Thanks for yoru results. Your distance curve is quite smooth apart from the two images that are out of focus. I think your results are better than mine. This could be down to your better tracking with the red dot sight or it could be because of a difference between our cameras.

I'm interested to know if your camera displays the same variability as mine. Would you mind shooting a burst of a small stationary object, one where the expected focus distance is unambiguous, using a single point. My focus distances vary quite a lot but yours may not. If that were the case, it would explain a lot of the diverse opinions in this thread.

Click to expand...

drj3 said:

Tony Rogers said:

Thanks for yoru results. Your distance curve is quite smooth apart from the two images that are out of focus. I think your results are better than mine. This could be down to your better tracking with the red dot sight or it could be because of a difference between our cameras.

I'm interested to know if your camera displays the same variability as mine. Would you mind shooting a burst of a small stationary object, one where the expected focus distance is unambiguous, using a single point. My focus distances vary quite a lot but yours may not. If that were the case, it would explain a lot of the diverse opinions in this thread.

Click to expand...

I posted a thread on focus accuracy of CAF vs SAF about 1 month ago. I used multiple single trials for that. There was some variability in focus which could be seen at 100% (table included in post), but that could be due to my test conditions. Most of the variability would not have affected any image viewed in a normal way.

I did find that CAF focused on the part with most contrast whereas SAF appeared to focus based on maximum overall contrast.

https://www.dpreview.com/forums/post/61338767

I haven't tested CAF focus on a stationary target with burst (though my feeding hummingbirds and slightly moving butterflies are close to stationary). I haven't noticed a problem since firmware 2.0 update. For that test I would want to use a very good tripod and set Image Stabilization to IS Priority if using the Mechanical shutter in case that caused sensor movement.

Click to expand...

Tony Rogers said:

drj3 said:

Tony Rogers said:

Thanks for yoru results. Your distance curve is quite smooth apart from the two images that are out of focus. I think your results are better than mine. This could be down to your better tracking with the red dot sight or it could be because of a difference between our cameras.

I'm interested to know if your camera displays the same variability as mine. Would you mind shooting a burst of a small stationary object, one where the expected focus distance is unambiguous, using a single point. My focus distances vary quite a lot but yours may not. If that were the case, it would explain a lot of the diverse opinions in this thread.

Click to expand...

I posted a thread on focus accuracy of CAF vs SAF about 1 month ago. I used multiple single trials for that. There was some variability in focus which could be seen at 100% (table included in post), but that could be due to my test conditions. Most of the variability would not have affected any image viewed in a normal way.

I did find that CAF focused on the part with most contrast whereas SAF appeared to focus based on maximum overall contrast.

https://www.dpreview.com/forums/post/61338767

I haven't tested CAF focus on a stationary target with burst (though my feeding hummingbirds and slightly moving butterflies are close to stationary). I haven't noticed a problem since firmware 2.0 update. For that test I would want to use a very good tripod and set Image Stabilization to IS Priority if using the Mechanical shutter in case that caused sensor movement.

Click to expand...

Thanks for the link. I hadn't seen that thread.

I would still be interested in your results for a static subject and C-AF/Burst. Might I suggest a relatively flat target like a single twig against the sky or at right angles to a wall. You target in the linked thread was looking obliquely at a the nails in the wooden boards which is what I meant by an ambiguous target. The camera could justifiably focus on different places within the focus point and get different results. My problem is the variability that I get when everything under the focus point is at the same distance but the camera still focuses at varying distances.

Click to expand...

GBC said:

I'm curious, is there parts of the photo that are tack sharp where it is in focus, or is the whole image soft.

Could be IBIS, not focus issues.

Click to expand...

Adrian Harris said:

GBC said:

I'm curious, is there parts of the photo that are tack sharp where it is in focus, or is the whole image soft.

Could be IBIS, not focus issues.

Click to expand...

An interesting point you make. I tested various m43 long lenses using stabilisation ON a few years ago, and got some weird results. I did actually write an article but upon checking I see I have never published it - probably because I was fed up with being attacked by fanboys every time I pointed out any camera/lens misbehavior!

Anyway the crux of the matter was when stood square on to a billboard and taking a series of shots with stabilisation on, the parts of the billboard that appeared to be in critical focus would vary from one place to another - even though I was square on - from shot to shot !!!

I could only put this down to lens optics or sensor 'wobbling' trying to stabilise large movements when hand holding a long zoom.

Anyway, since I no longer use IS for long zooms when shooting distant subjects I have greatly improved results.

Click to expand...