[MM] Changes in response to comments by Chris Catton

Dear all,

Chris Catton from the Image Bioinformatics Research Lab of University of 
Oxford sent me some very useful comments on the image annotation draft 
in a private mail, and gave me permission to forward them to this list:

>You talk too about Different vocabularies for different types of
>metadata. In our domain, we think it's important to go a step further
>and not just distinguish metadata about the properties of the image
>itself from those describing the subject matter of the image.  We also
>think it is important to distinguish 'ground facts' (I killed a rabbit,
>took out its liver, sectioned it and observed it under a microscope')
>from interpretation - 'The drug I gave the rabbit appears to have cured
>its cirrhosis'.  This is a difficult philosophical distinction, but one
>that is important to capture somehow.  We also see this as a separate
>issue from the use of domain-specific vocabularies for annotation.
>
>  
>
I fully agree, and I vaguely remember this distinction was made in 
earlier versions of the draft. Anyhow, I've rephrased the title to make 
it more generic (it now reads "Different types of metadata" iso 
"Different vocabularies for different types of metadata") and added the 
following sentence to that section:

     In many applications, it is also useful to distinguish between 
objective
    observations ('the person in the white shirt moves his arm from left 
to right')
    versus subjective interpretations ('the person seems to perform a 
martial arts exercise).

Note that I changed the example.  I thought Chris's example was a bit too
domain specific, but I 'm not too happy with this one either,  So I'm 
open for suggestions.

>We also emphasise the importance of capturing metadata at the point of
>creation (in our domain this often means extracting information from the
>microscopy image header files for example).  This reduces the burden on
>the researchers, reduces the cost of annotation, and reduces error.  I
>suggest this is something that would properly be addressed by a 'best
>practice' document and that at a quick glance appears to be missing from
>the current draft.
>
>  
>
Yes, this is, in my opinion, the number one best practice in multimedia 
annotation. 
I still cannot believe how we missed this in draft :-(
I added the following item as the first in the issues list in the intro 
section:

   1.

      Production versus post-production annotation

      Typically, most of the information that is needed for making the
      annotations is available during production time. Examples include
      time and date, lens settings and other EXIF metadata added to JPEG
      images by most digital cameras at the time a picture is taken,
      experimental data in scientific and medical images, information
      from scripts, story boards and edit decision lists in creative
      industry, etc. Indeed, maybe the single most best practice in
      image annotation is that in general, adding metadata during the
      production process is much cheaper and yields higher quality
      annotations than adding metadata in a later stage (such as by
      automatic analysis of the digital artifact or by manual
      post-production data).


>This all suggests to me that it is very much in our interest to work
>together and I would be happy to be involved.  The bioimage project is,
>I think, a very good real world application domain for this work - not
>least because it aims to deal with a wide variety of media types - not
>just stills and video. 
>

I asked Chris to try to find some example material from his project so 
we can put together a scientific use case from the bio imaging domain.  
I also added his name to the acknowledgement sections.

Chris, thanks again for your comments and I hope you will manage to find 
some example metadata for the use case.

Best regards,

Jacco

Received on Tuesday, 31 January 2006 13:13:25 UTC