Hello,

since I'm the author of PhotoME, and the one who decoded these tags, I think I can clean up some things. ;-)

I've called these matrices "Object Detection (_)?", because currently I don't know what they are used for, but to me they look like object detection informations- that where their names are coming from ;-)
The tag where these informations are saved is no invention by Sony, it is also available for Minolta cameras.

Like written above, I don't know what these matrices mean - but they seems to be important enough, so that Sony decided to write this stuff into the metadata.

There are some more matrices in this tag, which are currectly not shown by PhotoME.

I assume the matix which I've called "Object Detection (3)?" is somehow related to the auto focus, because while the other 2 matrices reflect the full image, "Object Detection (3)?" reflect only a smaller region in the middle of the image (the range where the AF points are located?).

I've also seen images where this matrix has no data (everything is zero) - maybe the auto focus was disabled for these images or the auto focus was not able to determine a focus point.

If someone has an idea what these matrices could be used for, or if someone has a better idea for the names of them - I would be really interested. ;-)

My hope is, that it would be somehow to possible, to determine the used AF point using these matrices.

Many regards,
Jens Duttke
Author of PhotoME

Very interesting stuff.

Carl

Jens,

Thank you for your clarification. Although your choice of "Object Detection" is very appropriate, another commonly used title for this type of technology is "Feature Detection" That is, the features of the image - luminance minima/maxima detection, edge detection, corner detection, keypoints, etc.

I have a few new questions. Besides Sony & Minolta, are there any other manufacturers that record this information? You mentioned other matrices, do the other matrices depict the scene in the image as clearly as these three? Or, do the other matrices seem to be more esoteric / are an even greater derivative of the image information?

Regarding your query about "what these matrices could be used for" - Are you asking why do these matrices exist in a DSLR image? Or, are you asking about this type of technology in general? That is, are you asking about uses such as Machine Vision, feature detection algorithms (e.g. SIFT used in Autopano), or tools like duplicate image finders?

Again thank you for your clarification.

You might like using exiftool, I use it often in little scripts. For example blast through a bunch of images in directories and see what my average focal length was etc.

It can read... a lot. Here's a dump from an A700 Raw basically telling it to dump all meta data even if you don't have any idea what it is.

Edited 4/21 to remove the HUGE exif list since Jens is capable of this as well - Eric

Hello,

"You might like using exiftool [...]
It can read... a lot. Here's a dump from an A700 Raw basically telling it to dump all meta data even if you don't have any idea what it is."
PhotoME will show you the same informations, and additionally to that the these matrices. :-)

Many regards,
Jens Duttke
Author of PhotoME

Hello,

"Besides Sony & Minolta, are there any other manufacturers that record this information?"

No, atleast I havn't seen something like that before.

"You mentioned other matrices, do the other matrices depict the scene in the image as clearly as these three? Or, do the other matrices seem to be more esoteric / are an even greater derivative of the image information?"

I havn't analysed the other matrices yet, if you are interested I'll try to decode the others at the end of this week.

Another interesting thing is maybe, that they all use different formats:
The first matrix has a size of 40x30 and 16 bits per value (I've seen values between 0x0 and 0x2FF)
The second matrix has a size of 40x30 and 8 bits splitted into 2 values (2 nibbles using values from 0 to 3)
The third matix has a size of 20x15 and 8 bit.

"Regarding your query about "what these matrices could be used for" - Are you asking why do these matrices exist in a DSLR image?"

Especially, I'm look for the correct name of them, so that users understand what they see there. :-)

Many regards,
Jens Duttke
Author of PhotoME

... I have used ExifTool for a couple of years but I didn't mention it because I thought that not everyone is comfortable with command line. Perhaps I should have mentioned ExifTool gui - It works well and rides over Exiftool. An other Exif reader that I like is ExifPro - it shows the Exif of a directory of images in columnar format - kinda' like that view that Sony's Picture Motion Browser provides.

Regarding ExifTool, note that some of those tags that ExifTool decodes is from Jens' input into ExifTool. That is, if you go on to the Exiftool history, Phil thanked various contributors and Jens seems to be a regular contributor to ExifTool, e.g. "Added values for various Sony tags (thanks Jens Duttke), Decode Nikon D300 custom settings (thanks Jens Duttke) . . . ". Then I noticed Phil called out Jens' work at the bottom of the ExifTool webpage - that's how I discovered PhotoME.

I've come to value Jens' work as I've used it to do various things - for example, re-inserting the original Exif info after some software program deletes it. Also, I've used it to "trick" DxO into thinking that my A700 image is from my A100. (Although all of that could be done with Exiftool, I just found it to be easier with PhotoME.)

Thank you for your response.

I haven't analysed the other matrices yet, if you are interested I'll try to decode the others at the end of this week.
Don't worry about decoding the other matrices - I can't use them - I only mentioned your matrices originally because I didn't know much about them.

However, if you are looking for features to develop ;-) then I have three specific feature request that others may find useful, too -
1) Could you allow editing for the lens tags (LensID & LensName B027). Why do I ask this? Because different manufactures use different LensID Schema and they can conflict with each other. Sometimes they are merely sister lenses, but sometimes they are different lenses altogether. For example, LensID 25531 is the ID for a Minolta 28-135 and it is the ID for a Sigma 28-105 - I would be hard pressed to say that those are sister lenses since they don't have the same range.

2) This 2nd feature may be more extensive - ExifPro (exifpro .com) allows users to view a list of images and their corresponding EXIF values in columnar format. That tool allows users to select which Exif values that would be shown in the columns. That is a particularly helpful view to determine which files get which sort of post processing treatment.

3) This last feature builds on the 2nd one I mention - if you could show that columnar view and and you combined it with your "quick lunch" then you *almost* become a central workflow tool. One of the other things that you would need to add for this feature, would be to allow the user to select a processing profile associated with a particular software package - The other - a much more complex feature for a latter release ;-) - would be to manage an image through multiple software packages

- - - - -

Another interesting thing is maybe, that they all use different formats:
The first matrix has a size of 40x30 and 16 bits per value (I've seen values between 0x0 and 0x2FF)
The second matrix has a size of 40x30 and 8 bits splitted into 2 values (2 nibbles using values from 0 to 3)
The third matix has a size of 20x15 and 8 bit.

Here are some of my on thoughts on *possible* reasons why they exist
1) Sony develops (and/or bought with the acquisition of Minolta's DSLR line) 'object detection / feature detection' algorithms used as "face detection", "smile detection" software in Point & Shoot (P&S) cameras.
2) Also, Sony develops both security cameras and some machine vision cameras that require object detection / feature detection.
3) In particular to Machine Vision cameras, there are a few different companies that produce software packages/algorithms to process the captured images from machine vision cameras, (National Instruments, MVTec, Dalsa-Coreco, etc).
4) Sony states that they provide data for these different algorithms. Each of those software packages *may* require different formats.
5) Sony regularly shares technology between it's different departments / organizations. (It's part of their effort to integrate and standardize Sony's internal organizational processes.)
6) These matrices *may* be remnants from the DSLR team borrowing software from the machine vision team and/or borrowing software from the P&S team.
--- So, all of these matrices *may* not be directly related to DSLRs - and *may* be just "extras" that are part of the code that would be used in Machine vision or in "face detection"

However, all of this is speculation. It would be interesting to find a Sony engineer willing to tell us. Albeit not very likely.

oh lol, he's a contributor. So he's at least one the very guys giving us support for some of the tags. Thank you Jens!

Great yes I love exiftool I use it for a lot of things in shell and perl scripts. I wish we'd see a bit more support for more exif in Lightroom and other programs for example it's really annoying. I wasn't sure if that was because of Sony was not divulging the information, or their plain lack of support.

I will have to give PhotoME a try.

thanks

Hello,

1) Could you allow editing for the lens tags (LensID & LensName B027). Why do I ask this? Because different manufactures use different LensID Schema and they can conflict with each other. Sometimes they are merely sister lenses, but sometimes they are different lenses altogether. For example, LensID 25531 is the ID for a Minolta 28-135 and it is the ID for a Sigma 28-105 - I would be hard pressed to say that those are sister lenses since they don't have the same range.

Can you please tell me the complete name of the Sigma lens? It is currently not in my database. If I would add it, PhotoME will show it correctly.

The Lens name cannot be changed, because it's based on the Lens ID. PhotoME has a look-up-table of (nearly) all lens ID's and the names of the lenses which use these IDs.

On the other hand, changing the Lens ID wouldn't make sense, if your lens isn't in the database at all.

If you tell me the name of your lens (or even better, send me an sample image), I'll add it to the database, so that PhotoME will detect it correctly.

2) This 2nd feature may be more extensive - ExifPro (exifpro .com) allows users to view a list of images and their corresponding EXIF values in columnar format. That tool allows users to select which Exif values that would be shown in the columns.

Version 0.8 will have a file browser with preview images. After this release I'll implement batch-processing functions, reporting functions and comparing functions, which base is this file browser, by and by.

3) This last feature builds on the 2nd one I mention - if you could show that columnar view and and you combined it with your "quick lunch" then you *almost* become a central workflow tool. One of the other things that you would need to add for this feature, would be to allow the user to select a processing profile associated with a particular software package - The other - a much more complex feature for a latter release ;-) - would be to manage an image through multiple software packages

That should be easily possible and will be part of the batch-processing functions.

Many regards,
Jens Duttke

Just thought I'd try out PhotoME, as well, and found it thought I'd used a Minolta AF 100mm F2.8 Macro (D) lens (LensID 28), although I actually used a Tamron AF SP90mm Di 2.8 macro. Just thought you'd like to know, Jens.

Cheers,
Richard.

Hello Richard,

thanks for this information!

New versions of the Konica/Minolta and Sony databases are now available on the PhotoME update server (in PhotoME: Settings-menu -> Online update...).
After the update your lens should be shown correctly.

Many regards,
Jens Duttke

Thanks - downloaded the update, and the Tamron lens is now shown. I then tried it out on photos from my Sony 50mm 2.8 macro (worked fine) & Sigma 70mm 2.8 macro (didn't work).

The Sigma lens is shown as Lens ID 25, "Minolta AF 100-300mm F4.5-5.6 (APO D) or Sigma 100-300mm F4 EX (APO (D) or D IF). I guess if you have logic which when looking up the Lens name from the ID, could look at the focal length used, and then determine that it's the Sigma 70mm 2.8 macro (as neither of the listed lenses go below 100mm) then it should be easy to correctly identify the lens. Shame the lens manufacturers don't use GUIDs :D

Oh - also checked PhotoME with my my old Sigma 28-70 F2.8 lens, and it shows as Lens ID 25511, Minolta AF 35-70mm F4 or Sigma 28-70mm F3.5-4.5 or Sigma 70-200mm F2.8 or Quantaray 35-80mm F-5.6. It's this lens, here: http://www.dyxum.com/lenses/detail.asp?IDLens=353

Hello Richard,

I guess if you have logic which when looking up the Lens name from the ID, could look at the focal length used...

PhotoME has such a logic - beside the lens ID it also consider the focal length and the max. aperture.
But this works only if all lenses are already in my database. ;-)
In the case of your Sigma 70mm f/2.8 Macro this isn't the case, so PhotoME will show all lenses using this lens ID.

A new update is now available which include your 2 lenses.

Shame the lens manufacturers don't use GUIDs :D

I assume it's a problem of the camera manufacturers.
Other camera manufacturers write much more informations into the metadata, which allow to detect the lens uniquely.

Many regards,
Jens Duttke

Looks good! Thanks mate!