ok..i just download the latest version from of A7info (v3). I went through a simple osprey mating and take off. All the eye are tacking sharp in the sequence final picture result. When using the Bird Eye AF, the box showing locking at the eye and zone dancing around when osprey turn the head.
When look for the focus point from A7info, it is not consistence showing some image is "animal eye tracking", some is "face tracking" and some is just plain "Zone" WITHIN the same sequence. The way i have it set up in the menu is "Face/eye AF priority" and turn on the bird eye on full time. So seem like during BIF, the camera will automatically pick whatever mode to grab face/eye or just regular focus area as long as the face/eye/head in focus.
Now, with a static osprey, the bird eye AF is more consistent showing at the eye or sometime to next to it (since the osprey have a black dot near his eye).
I'm upload the second set of images to google drive now. Just give you guy an update.
Alex Phan wrote:
ok..i just download the latest version from of A7info (v3). I went through a simple osprey mating and take off. All the eye are tacking sharp in the sequence final picture result. When using the Bird Eye AF, the box showing locking at the eye and zone dancing around when osprey turn the head.
When look for the focus point from A7info, it is not consistence showing some image is "animal eye tracking", some is "face tracking" and some is just plain "Zone" WITHIN the same sequence. The way i have it set up in the menu is "Face/eye AF priority" and turn on the bird eye on full time. So seem like during BIF, the camera will automatically pick whatever mode to grab face/eye or just regular focus area as long as the face/eye/head in focus.
Now, with a static osprey, the bird eye AF is more consistent showing at the eye or sometime to next to it (since the osprey have a black dot near his eye).
I'm upload the second set of images to google drive now. Just give you guy an update. ...Show more →
Thanks Alex!! Glad we have you on FM to give us the early sneak peaks as we all wait impatiently for the 4th.
For me, final result is what it matter and during the shot, it help to reduce my work flow (by not having to switch mode) and make it much easier, more enjoyable. That is what it matter. It get a passing grade for me. This will be the last time i upload file for this fun purpose. In my opinion, when compare to A9ii and R4, the eye/bird AF works much much better. It will grab the eye super quick and stick with it. It doesn't wonder around or jump to some high contrast area that is far away. Hit rate is more than reasonable. For those that want to play "prove it to me more", you can pound the sand.
I just look at the great blue heron file again, the hit rate got hit at the end because i zoom out the lens while panning/tracking from 309mm to 201mm.
actually the entire sequence are constantly zoom out to keep the subject fill the screen. I knew something is off that result in the out of focus. This is for sure is "user error".
For me, final result is what it matter and during the shot, it help to reduce my work flow (by not having to switch mode) and make it much easier, more enjoyable. That is what it matter. It get a passing grade for me. This will be the last time i upload file for this fun purpose. In my opinion, when compare to A9ii and R4, the eye/bird AF works much much better. It will grab the eye super quick and stick with it. It doesn't wonder around or jump to some high contrast area that is far away. Hit rate is more than reasonable. For those that want to play "prove it to me more", you can pound the sand. ...Show more →
Hardly surprising Tony N seemed to think it was a bit average - 5 frames out of 20 where it finds the eye. One frame on the eye, then next on the feet - interesting.
duncang wrote:
Hardly surprising Tony N seemed to think it was a bit average - 5 frames out of 20 where it finds the eye. One frame on the eye, then next on the feet - interesting.
How does TN know that? The sequences here and experience of many other tester showed that it works well. Even if he gets hardly any keepers, TN can't know whether it is a fail of BIF AF, or user error, or whatever, without a thorough analysis.
I am not aware of such an analysis, nor that he had the exact data from the camera indicating the actual mode used and focus point to make such a conclusion. So it is all speculation.
I don't know, too, how accurate the A7info software is. I have my doubts. I downloaded it and checked it with the A9. I didn't find the A7info software to give a reliable picture in many instances. The focus point was shown at various locations although the eye was sharp all the time. I am no convinced the software is fully reliable.
As Phan said, we only can rely on the end result and his experience (as is Dodds) is that it works very well.
For me, final result is what it matter and during the shot, it help to reduce my work flow (by not having to switch mode) and make it much easier, more enjoyable. That is what it matter. It get a passing grade for me. This will be the last time i upload file for this fun purpose. In my opinion, when compare to A9ii and R4, the eye/bird AF works much much better. It will grab the eye super quick and stick with it. It doesn't wonder around or jump to some high contrast area that is far away. Hit rate is more than reasonable. For those that want to play "prove it to me more", you can pound the sand. ...Show more →
Alex, thank you very much for posting these!
There is an enlightening part of the sequence as the bird takes off. If you look at 4309 through 4319, the eye is detected but appears to have missed the eye. However, the AF position reported is actually where the eye was in the previous frame (approx). Put your cursor on the eye in 4309 and then switch to 4310 to see it etc.
So it might seem like the camera is falling behind but this is mostly due to panning error. The camera has moved since when the focus position was calculated and AF enacted so the image is in focus even when the AF point seems to be completely off the bird. I think this phenomenon is also behind the complaints of cameras like the A7IV not having AF points on the subject during tracking but it is worse due to slower processor and frame rates.
Oh and my summary of the BEAF on these osprey examples is that it works very well!
somersettr wrote:
There is an enlightening part of the sequence as the bird takes off. If you look at 4309 through 4319, the eye is detected but appears to have missed the eye. However, the AF position reported is actually where the eye was in the previous frame (approx). Put your cursor on the eye in 4309 and then switch to 4310 to see it etc.
AF acquisition always happens with data taken before the actual shot. So if the target is moving across the sensor, the focus rectangle *should* show where the target was when the measurement took place, not where it happened to be by the time the shot was triggered.
So it might seem like the camera is falling behind but this is mostly due to panning error. The camera has moved since when the focus position was calculated and AF enacted so the image is in focus even when the AF point seems to be completely off the bird. I think this phenomenon is also behind the complaints of cameras like the A7IV not having AF points on the subject during tracking but it is worse due to slower processor and frame rates.
Yes and yes. This delay has been mentioned on multiple occasions, but apparently wasn't noticed by everyone interested. This makes it completely useless to analyse these AF positions on individual frames. You need to also look at the surrounding frames of a burst to check what the projected target speed on the sensor was.
This gets even more complicated when IS is on. But we all switch off IS when tracking birds, don't we?
duncang wrote:
Hardly surprising Tony N seemed to think it was a bit average - 5 frames out of 20 where it finds the eye. One frame on the eye, then next on the feet - interesting.
Since you seem to know: what was at the focus spot at the time the PDAF was measured? You did notice that IS was on?
Holger wrote:
How does TN know that? The sequences here and experience of many other tester showed that it works well. Even if he gets hardly any keepers, TN can't know whether it is a fail of BIF AF, or user error, or whatever, without a thorough analysis.
I am not aware of such an analysis, nor that he had the exact data from the camera indicating the actual mode used and focus point to make such a conclusion. So it is all speculation.
I don't know, too, how accurate the A7info software is. I have my doubts. I downloaded it and checked it with the A9. I didn't find the A7info software to give a reliable picture in many instances. The focus point was shown at various locations although the eye was sharp all the time. I am no convinced the software is fully reliable.
As Phan said, we only can rely on the end result and his experience (as is Dodds) is that it works very well....Show more →
I think A7Info is accurate only in that it correctly reads what is written to the file. If you zoom in on the rear of the camera set to zoom to focus point it will go to the same point registered in A7Info via the target graphic.
Once we have A1 in hand it will be easier to check as A1 has added the option to see the focus point in playback.
But, the info in the file is often not accurate as to what is really in focus. This has been a thing for ever on all camera systems that can show the focus point recorded. Half the time the point is where you wanted it and the subject is OOF and half the time the point is off where you wanted and yet the image is focused where you wanted. Of course once in awhile the things match up one way or the other
So yeah, I find little value in viewing these displayed points. A7Info is a little more interesting as it does show us what state the AF system was in for each shot (ie did it keep in RTT, did it find an eye, did it resort back to dancing dots).
Galer shared some now shots, a7info doesn't like the jpgs I downloaded from his flickr though. He said he had a very high hit rate (they were not in flight)
I think it is always good to put things in perspective. People that haven't shot an R5 may get the impression from reading posts that its Eye-AF is infallible. Well let me say it certainly fails in a lot of situations.
I'm certainly not saying the A1 is going to be better. But I was just watching this video from Duade and thought I'd share his short sequence showing the R5/6 fail miserably in Eye-AF.
Skip to the 3:27 mark for a short sequence demonstrating a failure...
buffalowolff wrote:
Galer shared some now shots, a7info doesn't like the jpgs I downloaded from his flickr though. He said he had a very high hit rate (they were not in flight)
multibit wrote:
Don't know if its just me but most of those the eyes head look OOF compared to the chest
I downloaded a few of the original files and checked them (several of them were the flying seagulls) in Photoshop. The eyes all look sharp at 100% magnification. The chest on the gulls were soft and there was not much feather detail, but that may have been due to blown highlights.
multibit wrote:
Don't know if its just me but most of those the eyes head look OOF compared to the chest
I just went through them all on Flickr by clicking into each image as far as it will let me. All looked right on the eye except maybe the 2nd Guiilame? image that may be sharpest on the body but the eyes still look pretty good. The one really close duck head shot has focus just starting on the eye and then backwards. Would have been better with the eye in the middle of the DOF or just a greater DOF for such a closeup shots.
The takeaway advice that I can offer at this moment in time is that the reliability of Bird Eye AF will vary between different shooting situations but any doubts I had about the 200-600 performing well with the A1 have now been removed.
