ExifTool reveals M10 metadata is.... sloppy work?

[TX400]

Advertisement (gone after registration)

I have been poking around a lot in the jpegs the M10 produces using Exiftool. I'm working on a project where I'm trying to get preview data out of the files as quickly as possible. Thus far I'm not impressed or at least confused by the data in the files.

 

Just about any operation throws out an ICC profile length warning, apparently when the declared profile length doesn't match the actual length of the profile:

Warning: Bad length ICC_Profile (length 3144)

 

Also try extracting the preview images and thumbnail images:

exiftool -a -b -W %d%f_%t%-c.%s -preview:all *.JPG

 

Extracted preview images are unreadable by chrome, safari, and preview.app, but oddly Firefox reads and displays and image. These are a nice size at 1440x940 and 133kb, and it'd be great to use them since the Maestro processor has already done the work to resize them.

 

The thumbnails are 160px wide and base64-encoded. They seem have black bars at the top and bottom. You can see them as iPhone iOS reads an SD card and otherwise on macOS. Extracting these works, but they have black bars along the top and bottom. Since I'm seeing this on the apple platforms and with ExifTool I think those bars may be part of the data.

 

The DNG is another story. It appears to be healthier. From the DNG, a similar thumbnail image is present with the black bars. There are two preview images embedded in the DNG file.  First is a ~1.6MB jpeg at 100/24MP size. Presumably this is the chimping image we actually see when we zoom in while using the back of the camera. The second is the 1440x940 small preview image from the jpeg, but this time it's intact and readable by usual programs.

 

It seems like something is wrong with the JPG files the camera creates, like the EXIF data is improperly copied or something. This is frustrating, for $7000 this should be correct.

Edited November 28, 2017 by TX400

[UliWer]

Sandymc found some „untidiness“ as well in his analysis of the raw files: https://www.l-camera-forum.com/topic/268507-leica-m10-raw-file-dng-analysis/?hl=exif

 

And there is even a hint by Mr. Daniel in an early interview about the M10 when he said they had wished to have more time to develop the M10.

 

Something to be „cleaned“ by a firmware update? One was expecting an update about the time whren they proposed the new CL though nothing seems to happen.

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[charles_nl]

Preamble: I am a layman as far as picture formats are concerned.  I used your 

 exiftool -a -b -W %d%f_%t%-c.%s -preview:all

command on a Leica JPG file and I get 2 files:

 

- L*_ThumbnailImage.jpg 160x120 (~1KB), which gives no problem to view with any Linux viewer I have (xv, gwenview, shotwell, gthumb, nomacs, pinta).  Has 2 black horizontal bars which cover part of the actual image with all viewers.

- L*_ThumbnailImage.jpgPreviewImage.dat (~200KB), apparently a data file that cannot be opened by a picture viewer.

- no 1440x940 preview image at all (but should there be one in a jpg file?).

 

The same exiftool command on the DNG file gives only 2 jpg files:

 

- L*_PreviewImage.jpg (~200KB) 1440x960 good picture, no black bars

- L*_JpgFromRaw.jpg (~1KB) 160x120 with the black bars (identical to L*_ThumbnailImage.jpg above)

- no additional ~1.6MB jpeg at 100/24MP size

 

I should note that my Leica JPG files have added meta data (IPTC).  The ICC warning is also always given

 

Charles

 

 

 

 

 

 

I have been poking around a lot in the jpegs the M10 produces using Exiftool. I'm working on a project where I'm trying to get preview data out of the files as quickly as possible. Thus far I'm not impressed or at least confused by the data in the files.

 

Just about any operation throws out an ICC profile length warning, apparently when the declared profile length doesn't match the actual length of the profile:

Warning: Bad length ICC_Profile (length 3144)

 

Also try extracting the preview images and thumbnail images:

exiftool -a -b -W %d%f_%t%-c.%s -preview:all *.JPG

 

Extracted preview images are unreadable by chrome, safari, and preview.app, but oddly Firefox reads and displays and image. These are a nice size at 1440x940 and 133kb, and it'd be great to use them since the Maestro processor has already done the work to resize them.

 

The thumbnails are 160px wide and base64-encoded. They seem have black bars at the top and bottom. You can see them as iPhone iOS reads an SD card and otherwise on macOS. Extracting these works, but they have black bars along the top and bottom. Since I'm seeing this on the apple platforms and with ExifTool I think those bars may be part of the data.

 

The DNG is another story. It appears to be healthier. From the DNG, a similar thumbnail image is present with the black bars. There are two preview images embedded in the DNG file.  First is a ~1.6MB jpeg at 100/24MP size. Presumably this is the chimping image we actually see when we zoom in while using the back of the camera. The second is the 1440x940 small preview image from the jpeg, but this time it's intact and readable by usual programs.

 

It seems like something is wrong with the JPG files the camera creates, like the EXIF data is improperly copied or something. This is frustrating, for $7000 this should be correct.

[TX400]

Thanks for running that Charles. 

 

I think the 200KB .dat file is supposed to be the jpeg PreviewImage. It looks like a similar problem to what I am having, but you got .dat and I got .jpg, both un-parsable by most viewers.

 

One thought I had is that the profile length being incorrect for jpegs is that they could be naively copying the DNG exif to the JPG, making the stated profile length different than the actual profile?!

 

In either case, this sounds like a firmware fix could solve the problem.

 

 

Preamble: I am a layman as far as picture formats are concerned.  I used your 

 exiftool -a -b -W %d%f_%t%-c.%s -preview:all

command on a Leica JPG file and I get 2 files:

 

- L*_ThumbnailImage.jpg 160x120 (~1KB), which gives no problem to view with any Linux viewer I have (xv, gwenview, shotwell, gthumb, nomacs, pinta).  Has 2 black horizontal bars which cover part of the actual image with all viewers.

- L*_ThumbnailImage.jpgPreviewImage.dat (~200KB), apparently a data file that cannot be opened by a picture viewer.

- no 1440x940 preview image at all (but should there be one in a jpg file?).

 

The same exiftool command on the DNG file gives only 2 jpg files:

 

- L*_PreviewImage.jpg (~200KB) 1440x960 good picture, no black bars

- L*_JpgFromRaw.jpg (~1KB) 160x120 with the black bars (identical to L*_ThumbnailImage.jpg above)

- no additional ~1.6MB jpeg at 100/24MP size

 

I should note that my Leica JPG files have added meta data (IPTC).  The ICC warning is also always given

 

Charles

[ramarren]

Olympus EXIF data is similarly weird and tweaky with EXIFtool, in every Olympus camera I've owned. Yet it works perfectly in Oly It's never once proved a difficulty to using the camera. I don't know why it matters.

[pico]

Olympus EXIF data is similarly weird and tweaky with EXIFtool, in every Olympus camera I've owned. Yet it works perfectly in Oly It's never once proved a difficulty to using the camera. I don't know why it matters.

 

Thoughtful engineering corrects what it can. Look at the hundreds of warning for most web pages, but they still function.

[ramarren]

Advertisement (gone after registration)

Thoughtful engineering corrects what it can. Look at the hundreds of warning for most web pages, but they still function.

 

 

Sure—fixing bugs and warnings is an important exercise for the engineer.

But if everything works, from a user perspective, why should the user be concerned about it? 

 

Remember that the error could also be in the EXIFtool parsing of the image file metadata...

Edited December 10, 2017 by ramarren

[pop]

Sure—fixing bugs and warnings is an important exercise for the engineer.

But if everything works, from a user perspective, why should the user be concerned about it? 

 

Remember that the error could also be in the EXIFtool parsing of the image file metadata...

Things that just happen to work might just happen to cease working at the smallest provocation. Let's say the user installs another version of the same software; or turns on or off an option which has no apparent connection to the fault in the file format; or uses another software product for no particular reason.

[pico]

Sure—fixing bugs and warnings is an important exercise for the engineer.

But if everything works, from a user perspective, why should the user be concerned about it?

 

In the end it is about social order.

 

Godfrey, some of us non-engineers hope technology works in a concerted effort to make determinable and efficient works. The so-called Web languages are a mess; a huge patch-work of a technology that highlights the methodological language errors we should despise if they were, for example, evinced in public policy.

 

Aside: I follow deep learning and related research and it is interesting that what appears to work sometimes does so without the rigor of science.

Edited December 10, 2017 by pico

[Bill W]

This is giving me a headache, think I will take my M10 out and take some images....!!

[anickpick]

 

Aside: I follow deep learning and related research and it is interesting that what appears to work sometimes does so without the rigor of science.

 

Sometimes? ... less than 30 percent of western medical practice is based on valid scientific evidence

[pgk]

Things that just happen to work might just happen to cease working at the smallest provocation.

 

In that case we are all in real trouble. Depends what 'just' makes them work. And provocation is, well, provocation. Currently the files work in the way we normally utilise them - in the future, who knows(?) and the reasons may or may not be linked to 'sloppiness'. 

[TX400]

I just tested on the new 1.10.4.0 firmware and the same problem exists.