From: Cline, Danelle
Sent: Wednesday, August 31, 2005 12:16 PM
To: Schlining, Brian; Cline, Danelle
Cc: Chase, Andrew
Subject: RE: AVED classifier requirements doc update and project share

Thanks for the clarification Brian. 

>>  In fact, I suspect that we will be using AVED to identify down to genus or family more often than species. For example, Nanomia is a genus, and I feel that resolution to species, such as Nanomia bijuga or Nanomia cara, isn't likely to occur using AVED in the near future.

Yes, I agree. I stand corrected.  That's what you get when you ask and engineer a question about species :) Duhh.

>>Class (actually ClassImageSet) can not be deleted once they are locked (i.e. the Locked bit is toggled in the database)


>>We'll need to consider how and when we allow Images to be deleted. 

If images in a class are not locked and not referenced elsewhere, they should be deleted.  This means physically deleting the images on Nanomia's RAID and the reference in the DB. If this means images should not be indestructible, then we should change this. 

This allows a user to delete an experimental class which will be needed in the near-term.  Also, as technology advances we may, for example, replace our SD images with larger HD images, and we could run into storage issues down the road on our RAID if we can't free up some space. 

-Danelle


--------------------------------------------------------------------------------

From: Brian Schlining [mailto:brian@mbari.org] 
Sent: Wednesday, August 31, 2005 11:33 AM
To: Cline, Danelle
Cc: Chase, Andrew
Subject: Re: AVED classifier requirements doc update and project share


      page 4, Definitions
      "A class shall be defined as a collection of related images that
      represent a particular animal, collection of  animals, or objects."
      does "collection of animals" == "collection of animals of the same
      species"? 




  yes.


Just my 2 cents...A class, or ClassImageSet as it's called in the AVED database, is a collection of images representing a particular concept. [Where a concept is usually defined in the VARS knowledge-base.] The concept may be a species, but it can also be more generic. In fact, I suspect that we will be using AVED to identify down to genus or family more often than species. For example, Nanomia is a genus, and I feel that resolution to species, such as Nanomia bijuga or Nanomia cara, isn't likely to occur using AVED in the near future.


      Are new classes restricted to what's in the
      VARS KB? Or will the looser definition stay in place, that classes


  may


      come from the VARS KB (which i read as "may not" as well).




  Yes, classes will typically come from the VARS KB, but we also may
  test things outside our KB for collaborations. I'll update the reqt 
  doc to reflect this as it's not clear.




From the database side of things this is fine. A ClassImageSet has a ConceptName field which is just a string. We can put anything we like in that field.


  Deleting classes and images:
  user: can delete an entire class
  backend: call to database to delete class.
  user: delete images within a class
  backend: delete references to image in class model.


Class (actually ClassImageSet) can not be deleted once they are locked (i.e. the Locked bit is toggled in the database)


Images are considered immutable and indestructible, they may also be shared between different ClassImageSets. We'll need to consider how and when we allow Images to be deleted. The KISS rule would say that once they are committed to the database they can not be deleted as long as any ClassImageSet references them.


B

~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
Brian Schlining
MBARI
Software Engineer
brian@mbari.org
(831)775-1855
http://www.mbari.org/staff/brian