Using “official” Exif schema as an alternative of extended image metadata?

[wrznr]

After playing around with the extended PAGE image metadata fields, @Boenig and I wanted to suggest further additions. However, this would ultimately lead to an inclusion of a complete image metadata set into PAGE which might not be desirable. Relying on an already existing XML schema is probably an effective alternative. Luckily, such a schema is existing and, even better, it is maintained under the tutelage of the W3C: https://www.w3.org/2003/12/exif/

So, what do @chris1010010, @splet and @cneud think about simply including this rdf schema in PAGE XML?

[chris1010010]

Being in line with exif makes sense. A few problems:

• We can't just integrate an RDF schema (we need an XML schema and I couldn't find one quickly)
• We only need a few attributes from exif
  Can we discuss quickly at some point? What is useful etc. And why these attributes should be in PAGE (and not just in the linked image). Some redundancies are helpful (such as the width and height in order to make sure the coordinates of page content matches the image). But the more redundancy we add the bigger the risk of mismatches (PAGE says one thing and the image metadata says something else).

[chris1010010]

Hi, So I have an alternative proposal. - We add resolution and resolutionUnit (because it’s crucial information to convert from pixel coordinates to mm etc.) - We add a new concept of generic MetadataItem elements o Each element can point to an external model/schema/ontology That way, we don’t make PAGE too specific and can use the EXIF schema as suggested. The interpretation/validation is then shifted to the application creating/reading the PAGE file. Best, Christian From: Kay-Michael Würzner <[email protected]> Sent: 05 January 2018 12:24 To: PRImA-Research-Lab/PAGE-XML <[email protected]> Cc: Christian Clausner <[email protected]>; Mention <[email protected]> Subject: [PRImA-Research-Lab/PAGE-XML] Using “official” Exif schema as an alternative of extended image metadata? (#3) After playing around with the extended PAGE image metadata fields, @Boenig <https://github.com/boenig> and I wanted to suggest further additions. However, this would ultimately lead to an inclusion of a complete image metadata set into PAGE which might not be desirable. Relying on an already existing XML schema is probably an effective alternative. Luckily, such a schema is existing and, even better, it is maintained under the tutelage of the W3C: https://www.w3.org/2003/12/exif/ So, what do @chris1010010 <https://github.com/chris1010010> , @splet <https://github.com/splet> and @cneud <https://github.com/cneud> think about simply including this rdf schema in PAGE XML? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#3> , or mute the thread <https://github.com/notifications/unsubscribe-auth/AHgOxrhOZBivsMScKRTh-jDMVf4KqZidks5tHhRigaJpZM4RUYmL> . <https://github.com/notifications/beacon/AHgOxhKI70r7CGiF-m_VQXbTjWyNFx7Jks5tHhRigaJpZM4RUYmL.gif>

[kba]

Our current understanding at @OCR-D is to not encode EXIF metadata as part of PAGE at all (except for imageWidth/imageHeight for convenience/compliance) but to require image files to contain EXIF data such as pixel density or color depth.

IMHO it's great to have a generic metadata mechanism and attribute for encoding pixel density in PAGE, esp. for ALTO-PAGE interop but from @OCR-D's perspective this issue is resolved.

@wrznr Correct me if I'm wrong.

[chris1010010]

Agreed. Closing :-)

[bertsky]

Our current understanding at @OCR-D is to not encode EXIF metadata as part of PAGE at all (except for imageWidth/imageHeight for convenience/compliance) but to require image files to contain EXIF data such as pixel density or color depth.

In the meantime, that understanding has been supplanted by the opposite view – for OCR-D. So it's precisely because EXIF/JFIF/... metadata are so unreliable in practice that the additions here in PAGE are vital. We need to (be able to) add metadata from an external/automatic image characterization to the annotation, without changing/copying the original image.

However, I still don't quite understand how the new MetadataItem can be used to host information in the EXIF schema, without using an external schema directly.

@chris1010010, you added some documentation, but no example. Can you please elaborate?

Let's take https://www.w3.org/2003/12/exif/#example – in RDF this looks like this:

<exif:IFD rdf:ID="Primary_Image">
   <exif:make>Canon</exif:make>
   <exif:model>Canon IXY DIGITAL 30</exif:model>
   <exif:orientation>top-left</exif:orientation>
   <exif:xResolution>180/1</exif:xResolution>
   <exif:yResolution>180/1</exif:yResolution>
   <exif:resolutionUnit>inch</exif:resolutionUnit>
   <exif:dateTime>2003-01-18T16:07:30</exif:dateTime>
   <exif:yCbCrPositioning>centered</exif:yCbCrPositioning>
   <exif:exif_IFD_Pointer>
    <exif:IFD>
      <exif:exposureTime>1/400</exif:exposureTime>
      <exif:fNumber>35/10</exif:fNumber>
      <exif:exifVersion>2.20</exif:exifVersion>
      <exif:dateTimeOriginal>2003-01-18T16:07:30</exif:dateTimeOriginal>
      <exif:dateTimeDigitized>2003-01-18T16:07:30</exif:dateTimeDigitized>
      <exif:componentsConfiguration>01 02 03 00</exif:componentsConfiguration>
      <exif:compressedBitsPerPixel>5/1</exif:compressedBitsPerPixel>
      <exif:shutterSpeedValue>76</exif:shutterSpeedValue>
      <exif:apertureValue>116/32</exif:apertureValue>
      <exif:exposureBiasValue>96</exif:exposureBiasValue>
      <exif:maxApertureValue>116/32</exif:maxApertureValue>
      <exif:meteringMode>Pattern</exif:meteringMode>
      <exif:flash>Flash fired, compulsory flash mode, red-eye reduction mode</exif:flash>
      <exif:focalLength>294/32</exif:focalLength>
      <exif:flashpixVersion>1.00</exif:flashpixVersion>
      <exif:colorSpace>sRGB</exif:colorSpace>
    </exif:IFD>
   </exif:exif_IFD_Pointer>
 </exif:IFD>

We can see the recursive nature of IFD at work here. How do we format that using the strict MetadataItem/Labels/Label hierarchy? By flattening? And what do we put in MetadataItem/@name and @value? What in Labels/@externalModel and @externalId?

Is the following the correct/intended representation for this example?

        <pc:MetadataItem type="imageProperties" name="?" value="?">
          <pc:Labels externalModel="http://www.w3.org/2003/12/exif/ns" externalId="IFD" prefix="exif">
            <pc:Label value="Canon" type="make"/>
            <pc:Label value="Canon IXY DIGITAL 30" type="model"/>
            <pc:Label value="top-left" type="orientation"/>
            <pc:Label value="180/1" type="xResolution"/>
            <pc:Label value="180/1" type="yResolution"/>
            <pc:Label value="inch" type="resolutionUnit"/>
            <pc:Label value="2003-01-18T16:07:30" type="dateTime"/>
            <pc:Label value="centered" type="yCbCrPositioning"/>
          </pc:Labels>
          <pc:Labels externalModel="http://www.w3.org/2003/12/exif/ns" externalId="exif_IFD_Pointer" prefix="exif">
            <pc:Label value="1/400" type="exposureTime"/>
            <pc:Label value="35/10" type="fNumber"/>
            <pc:Label value="2.20" type="exifVersion"/>
            <pc:Label value="2003-01-18T16:07:30" type="dateTimeOriginal"/>
            <pc:Label value="2003-01-18T16:07:30" type="dateTimeDigitized"/>
            <pc:Label value="01 02 03 00" type="componentsConfiguration"/>
            <pc:Label value="5/1" type="compressedBitsPerPixel"/>
            <pc:Label value="76" type="shutterSpeedValue"/>
            <pc:Label value="116/32" type="apertureValue"/>
            <pc:Label value="96" type="exposureBiasValue"/>
            <pc:Label value="116/32" type="maxApertureValue"/>
            <pc:Label value="Pattern" type="meteringMode"/>
            <pc:Label value="Flash fired, compulsory flash mode, red-eye reduction mode" type="flash"/>
            <pc:Label value="294/32" type="focalLength"/>
            <pc:Label value="1.00" type="flashpixVersion"/>
            <pc:Label value="sRGB" type="colorSpace"/>
          </pc:Labels>
        </pc:MetadataItem>

[chris1010010]

I'll be out of the office until January.
As far as I remember the idea was to use the Label mechanism to add semantic information (e.g. exif tag types) to the MetadataItem instances.
Edit: just noticed that that's what you did. So, yes, that's one option to realise it

[bertsky]

This can wait. Thanks @chris1010010 !