Uploaded image for project: 'Suggestion Box'
  1. Suggestion Box
  2. SUGGEST-121

Machine-based labeling of assets in DAM


    • Type: Bug
    • Status: Closed
    • Priority: Neutral
    • Resolution: Duplicate
    • Labels:


      With the ever increasing availability of external services that allow automatic labeling/annotation of resources it might be quite useful to have access to these services also in Magnolia, for example for the resources uploaded in the Assets app.

      While not limiting this idea only to images, the images being uploaded in the Assets app is the first thing that comes to mind. The upload dialog could feature a button, which would - send a labeling/annotation request for the given resource (image in this case):

      If processed successfully by the external recognition/labeling service, the result would be inserted into a new "Labels" field in the dialog:

      The labels would be saved upon clicking "SAVE CHANGES".

      The assets browser could have a new "Labels" column which would show the labels associated with the given resource:

      In the configuration of the DAM module the user could configure whether the generation of labels should be done automatically with each upload action. The user could also configure a labeling threshold/limit (e.g. 1000 images), which - if reached in the instance for the defined period of time (e.g. month) - would impose a ban on the labeling function and inform the user about it. The user could then consider upgrading the plan subscribed to with the labeling service.

      Example configuration

      An extension to the Content Tags module

      The module could use the information stored in the form of asset labels to automatically generate a first content-tag suggestion for pages an author could either accept, decline or modify, provided of course a page contains labeled assets (images). To do this the DAM labeler should probably store in JCR also probability for each label generated.

      Further reading


          Issue Links



              • Assignee:
                mdrapela Martin Drápela
              • Votes:
                1 Vote for this issue
                1 Start watching this issue


                • Created: