[licht96]

Wow... What a detailed view on the best settings for every situation for this Nikon. That is really missing for new Pentax.
May be after a year or two we have something like this. And after final software updates, too.

[clackers]

Sometimes call subject recognition (e.g. K-3 II), often subject detection. It is the same thing.

It is not. In the old cameras, the pattern under the focus point is registered and in real time, if that pattern appears under an adjacent point, the focus is shifted to that point.

In the K-3 III, the whole picture or an area of the picture is examined for the camera recognising a face or a bird or whatever. The focus will jump to an eye, for example.

Your K-3 II just cannot do this, Ulrich, it has one third the precision!

What the K-3 III does is similar to what the Sony A9 II I used in March this year does.

[UlrichSchiegg]

[...] Sometimes call subject recognition (e.g. K-3 II), often subject detection. It is the same thing. Your statement, that the 86k Sensor resolution is too low does origin from where? [...]

It is not. In the old cameras, the pattern under the focus point is registered and in real time, if that pattern appears under an adjacent point, the focus is shifted to that point.

In the K-3 III, the whole picture or an area of the picture is examined for the camera recognising a face or a bird or whatever. The focus will jump to an eye, for example.

Your K-3 II just cannot do this, Ulrich, it has one third the precision!

What the K-3 III does is similar to what the Sony A9 II I used in March this year does.

I don't have a K-3 II. I have the K-S2 with the old type of metering sensor (77-Segment) and the KP which also has the 86K-Pixel-Type AE-Sensor. And now the K-3 III.

Whether it is, or whether it is not, whether the 86k sensor has too small resolution or has not, is a pure speculative thing, without a reference to some type of Pentax documentation. The only type of documentation that I know is this "Pentax Real Time Analysis System"-graphs that Ricoh releases under the feature descriptions of the cameras. That hasn't changed since its 1st appearance for the K-3 and is now also included for the K-3 III. And similar pictures exist for Nikon and Canon DSLRs.

Phowon (Reference below) said that he has been testing the K-3 III AF with Ricoh. He gave some indication on "Trained Modeling Data" Software used. Whether it is this one, an inhouse version or somerthing else, doesn't really matter. It indicates well enough how these systems work. I don't think that there is something fundamentally different for the K-3 III. But who knows. Regarding tracking initialisation, it may well be that for only some occasion the 86k-Sensor was good enough for detection and for many others situations it didn't resolve "enough". It may also be that the old processing hardware was too slow for many cases. So the initialization worked, but the tracking was too slow, and therefore the camera had to rely mostly on the AF-Sensor for tracking. Usually these methods don't come in black and white.

What I fully agree to is, that the K-3 III does it much better. I have it for more than two weeks now and have been to some sports events. The 1st thing that did strike me is this huge view finder magnification compared to the KP, which allows much more intermediate action with the sports. Initially I thought that I am zooming too much, until I realized it is the magnification, that's different. The joystick together with the speed that camera reacts to changes (locking of focus etc.) makes it much more responsive to user interaction while looking through the view finder. The different AF modii I am still experimenting with (zone versus expanded area). But I mostly stick with the Nikon and Canon recommenations, since there are none published by Ricoh. And the tracking and AF precision is just great, compared to the KP.

Umm...

Deep learning is just for "Firmware development" not YOUR each Camera.

OpenCV: Cascade Classifier
OpenCV: samples/cpp/facedetect.cpp

for Face detection and Eye-AF, a Trained modeling data is needed. This data is already included in firmware binary file. (and will be any datafile of filesystem in DSLR Flash Memory).

and detection performance(speed,accuracy...) is depend on that modeling data is Well-trained or not.
Trained modeling data is made by (psudo code: ) train(many many face photos) method.

Mr.Wakashiro, Chief of Dev.team for PENTAX K-3 Mark III said "if needed, any firmware may be made for specific purpose like for airplane detection... " on CP+2021. This means same to my opinion.

btw, I'm sure PENTAX may use OpenCV by SC2000, just my think. lol
(Don't trust. PENTAX never officially said "we use SC2000". lol)

[clackers]

Whether it is, or whether it is not, whether the 86k sensor has too small resolution or has not, is a pure speculative thing

?

There's no speculation, Ulrich.

The K-3 III has more than 300,00 pixels in its sensor now.

Check out all the videos (like Kobie has made) for focus jumping to the eye, for example.

Those capabilities are completely new, but you must obey the documentation.

[UlrichSchiegg]

[...] The subject recognition I am talking about and the Zone modes it needs do not exist in the older K-3. The resolution is too low. The new K-3 III triples the number of pixels, so it can do eye and bird focus in the OVF, for example. You cannot do that with the K-3.[...]

[...] Sometimes call subject recognition (e.g. K-3 II), often subject detection. It is the same thing. Your statement, that the 86k Sensor resolution is too low does origin from where? [...]

It is not. In the old cameras, the pattern under the focus point is registered and in real time, if that pattern appears under an adjacent point, the focus is shifted to that point. In the K-3 III, the whole picture or an area of the picture is examined for the camera recognising a face or a bird or whatever. The focus will jump to an eye, for example. Your K-3 II just cannot do this, Ulrich, it has one third the precision! [...]

Whether it is, or whether it is not, whether the 86k sensor has too small resolution or has not, is a pure speculative thing, without a reference to some type of Pentax documentation. The only type of documentation that I know is this "Pentax Real Time Analysis System"-graphs that Ricoh releases under the feature descriptions of the cameras. That hasn't changed since its 1st appearance for the K-3 and is now also included for the K-3 III.

?

There's no speculation, Ulrich.

The K-3 III has more than 300,00 pixels in its sensor now.

Check out all the videos (like Kobie has made) for focus jumping to the eye, for example.

Those capabilities are completely new, but you must obey the documentation.

Don't understand your question mark. You made statements that the 86k AE sensor has not enough resolution and that the "detecion/recognition" doesn't works for older cameras with the 86k-Sensor. And that for the K-3 III the whole image is analysed while for the older ones there is some type of pattern analysis. I still don't understand the origin of that statement, that's why I called it speculation.

The Pentax/Ricoh documentation doesn't indicate so. There is little documentation available. One is that Video:

You Tube

You Tube

"We also revised the algorithm for the approximate calculation, to further improve the accuracy of focus prediction", "the camera expands the range of scenes detectable by deep learning". To me that indicates evolution instead of revolution.

Kobe videos? Are we at this level? Did anybody try the KP or K-1 in full auto mode sloooow motion head turning. You want to use full Auto mode AF for sports?

Anyway, since there is no documentation this gets into a right or wrong discussion, in which I have no interest. I may well be wrong.

Take care, always enyoing your opinion clackers.

[Robert N]

I also suffered with the 'blank screen' scenario. It happens if you push the 'info' button twice whilst viewing the status screen/control panel. Push the 'info' button again to restore the status screen. I hope this is helpful.

[Kobie]

Don't understand your question mark. You made statements that the 86k AE sensor has not enough resolution and that the "detecion/recognition" doesn't works for older cameras with the 86k-Sensor. And that for the K-3 III the whole image is analysed while for the older ones there is some type of pattern analysis. I still don't understand the origin of that statement, that's why I called it speculation.

The Pentax/Ricoh documentation doesn't indicate so. There is little documentation available. One is that Video: PENTAX K-3 Mark III [Challengers] IV. Image tracking performance - YouTube "We also revised the algorithm for the approximate calculation, to further improve the accuracy of focus prediction", "the camera expands the range of scenes detectable by deep learning". To me that indicates evolution instead of revolution.

Kobe videos? Are we at this level? Did anybody try the KP or K-1 in full auto mode sloooow motion head turning. You want to use full Auto mode AF for sports?

Anyway, since there is no documentation this gets into a right or wrong discussion, in which I have no interest. I may well be wrong.

Take care, always enyoing your opinion clackers.

Yes I have used the K-3 in full auto AF to see if there's any subject recognition regarding eye tracking AF and there is not. The KP uses the same system as the K-3.
The difference here with the K-3 Mark III is the addition of both the new RGBir sensor with triple the resolution that can be tied directly to the AF system AND the deep learning algorithm which no previous Pentax camera had.

Without getting into crazy technical talk, Pentax has explained all this and if you understand the tech (what it is and how it works) the manual does give you insight regarding the options.
"AF Type 1 combines the RGBir sensor with the AF sensor for more accurate tracking"
"Subject recognition only works in Auto or Zone AF modes".
Subject recognition allows for eye tracking AF through the OVF (Stated in the Pentax pre-release videos).
All the information is there. No other Pentax cameras have had this capability.

[UlrichSchiegg]

I do not doubt the K-3 III. For sports I use similar approaches that clackers does (expanded area etc.) or as are explained in the D500 docs. I doubt the statements about the earlier models like KP (the one I have) and that there is something fundamentally different. For the K-3 III everything is better (resolution, speed etc.). The KP had a very usable expanded area mode for sports. It is with the K-3 III that I feel everything is quicker, better, more resonsive etc. But fundamentally different? No.

PS Color tracking was definitely there with the KP. Hit rate is different depending what shirts the teams did wear (red best). So this real time analysis system was definitly there with the KP. The detection/recognition thing, I am not sure that this is not just a marketing gimmick (as with the mirrorless). For Sports this Auto mode doesn't help. With fast lenses I rarely had a problem getting the eye properly focused with the KP and the expanded area mode. K-3 and KP don't have the same system. K-3: 2013 SAFOX 11, PRIME IV KP: 2017 SAFOX 11, PRIME III. The AE sensors 86K are the same.

Anyway, as I wrote this discussion appears to me as not of any benefit. At the next sports event, I will try the zone mode. With the KP I didn't use it. Neither Canon nore Nikon recommend it for sports.

To get back to the thread topic. The settings I use are similar to yours Thor. For the tracking initialization, the focus priority is important. On the KP I had the 2+ images on release priority. With the speed of the K-3 III I may switch this to focus priority as well. Will try. For metering, I switch depending situation. Hold I put depending sports. With Sports I try and not photograph against the sun, so light situation is rather constant. Then it doesn't matter if TV, Av or TAv is used. AutoIso I rarely use.

[beholder3]

"Subject recognition only works in Auto or Zone AF modes".

I still find it amusing that this sentence has to be written down. It is like "beware, water is wet".

If one understands that "subject recognition" is used to follow a pattern and then automatically activate/use AF-Points withing the boundaries of that recognized subject outline it is absolutely clear that this can only work in a mode where you have allowed the camera to choose from a very wide array of AF points.

Thinking the other way: What if you restricted the AF points to "single select" and put that in the upper right corner while subject recognition is on and detects a face in the left side. Basically the resulting thoughts of the camera would be "Ok, master, yes, I have detected a face, but no, I will not focus there as you have explicitly told me to only use that single tiny AF point which you have decided to put into the upper right corner."

Another comment:
My good old K-1 I already has some basic pattern recognition based on the metering sensor when using the OVF. It is not great and processing power is very slow, but it is there.
It's easy to prove. Select "large select" AF area in AF.C and point the selected AF-sensor on the eye of a face with back AF-button pressed. Then very (!) slowly swivel the camera. The active AF point will stick (more or less) to the eye or more exactly the camera will activate different AF points to where it detects the eye.
The better the software for pattern recognition and the more processing speed you have, the better it works.

---------- Post added 23rd Sep 2021 at 09:10 ----------

It is not. In the old cameras, the pattern under the focus point is registered and in real time, if that pattern appears under an adjacent point, the focus is shifted to that point.

I do have to object here (based at least on my understanding).

The AF sensor only see greyscale pixel lines and as such are completely unable to recognize anything I'd call a "pattern". They basically only identify they biggest jump in brightness (=contrast edge) in that pixel line.
The data they have is along the following for one line AF point which is actually made up out of two separate physical sensor lines:
1: 222222224888888
2: 222488888888888
The numbers representing brightness values.
The contrasty edge is where the values go from 2 over 4 to 8 and the focus is moved so that both sensor detections match.

Old style handover between multiple AF points to my understanding was just based on the assumption that the desired subject is always the closed thing as unwanted background is farther back. So basically the camera activated any AF points which were similarly "close" to the initial/previously activated AF points(s) reading when the initial AF point suddenly lost it (= the distance suddenly was much bigger).
Or in other words: The camera always had a super simplistic 3D map from the AF sensor readings (consisting of say 11 data points of depth data) and chose the AF point where the "closeness" peaked.

[clackers]

The AF sensor only see greyscale pixel lines and as such are completely unable to recognize anything I'd call a "pattern".

Yes, but it's not the AF sensor that identifies the pattern. It's the RGB metering sensor that does that, with 307,000 pixels in the K-3 III instead of 86,000 in previous cameras.

For sports and action shooting today, the K-3 III is *the* Pentax of choice.

[dicki]

and why does the focuspoint jump to another point when iI had focus it?
Really. Your wourkflow is right with F11, but when I focus with small area in AF-C the point will jump in other way, always.
(My englich ist bad, I know!)

Here is a video from a man who want to shoot a Komoran with AF-C. Is does not work.

With the K3 I had one point in one area. That works!
Maybe I need your help.

You Tube

You Tube

[Kobie]

and why does the focuspoint jump to another point when iI had focus it?
Really. Your wourkflow is right with F11, but when I focus with small area in AF-C the point will jump in other way, always.
(My englich ist bad, I know!)

Here is a video from a man who want to shoot a Komoran with AF-C. Is does not work.

With the K3 I had one point in one area. That works!
Maybe I need your help.

Pentax K-3 III AF.S and AF.C AF modes demonstrated, and when to use which one. It ain't like it was. - YouTube

From all my shooting and testing, here's what I've found.
This is a very exacting camera. If you're shooting a still subject, you MUST use AF.S.
AF.C seems to look specifically for anything that's moving, including ripples of water.
Or you can switch the AF type in AF.C to Spot, SEL, or Sel-s so it only triggers a single AF point.
For birds in flight or other moving subjects with a clear static background, you can use Zone or Auto AF in AF.C and it'll be fine.
The K-3 Mark III just doesn't behave the way any of the previous Pentax cameras did, so you have to "forget" everything you used to do and learn to use all the settings on this camera because what you used to do won't work on the K-3 Mark III. It's a totally different experience and takes time to figure out.

[dicki]

From all my shooting and testing, here's what I've found.
This is a very exacting camera. If you're shooting a still subject, you MUST use AF.S.
AF.C seems to look specifically for anything that's moving, including ripples of water.
Or you can switch the AF type in AF.C to Spot, SEL, or Sel-s so it only triggers a single AF point.

I don't know what you mean. On SEL-S it will be triggers a single AF-Point in a field of 5 points. Have you find a way that does not?


For birds in flight or other moving subjects with a clear static background, you can use Zone or Auto AF in AF.C and it'll be fine.
The K-3 Mark III just doesn't behave the way any of the previous Pentax cameras did, so you have to "forget" everything you used to do and learn to use all the settings on this camera because what you used to do won't work on the K-3 Mark III. It's a totally different experience and takes time to figure out.

Yes I know, it's take time to figure out. And it is allways new. But what can I do if the point will be change? I don't want shoot with F16 always.
I need more help, please.

[normhead]

Great thread guys.... I just hope I get to use the information some day.

[Kobie]

Yes I know, it's take time to figure out. And it is allways new. But what can I do if the point will be change? I don't want shoot with F16 always.
I need more help, please.

As I mentioned...
This is a very exacting camera. If you're shooting a still subject, you MUST use AF.S.
AF.C seems to look specifically for anything that's moving, including ripples of water.
Or you can switch the AF type in AF.C to Spot, SEL, or Sel-s so it only triggers a single AF point.
For birds in flight or other moving subjects with a clear static background, you can use Zone or Auto AF in AF.C and it'll be fine.
Use the AF types (spot, select, select small, zone and full auto) where appropriate. You need to be specific with this camera. If you want to shoot a subject when using Zone or SEL-L/M or full auto and the camera is picking other targets, then change the focus type to SEL-S, Select, or spot.