Usage of the plugin

We developed the plugin with the idea in mind that it should not change the way a user is interacting with the Cumulus DAM system. This way it is very easy to take advantage of the features the plugin provides.

The feature set basically consists of different types of automatic recognition. Every feature enabled consumes one request available in your Clarifai account.

To decide which features you want to use you have record fields available in your Cumulus catalog to control it if the catalog template was imported like described in the Installation.

Fields for controlling the used features of the plugin:

  1. AutoTag_Model (list field): defines which public tagging model should be used (default: "General") or which custom models to use (see section 5. in this chapter)
  2. AutoTagging Color (boolean field): decides weather to use the recognition of the dominant colors or not (default: true)
  3. AutoTagging Classification (boolean field): decides use the recognition of "nsfw" content or not (default: true)
  4. AutoTagging Face (boolean field): decides wheter to use the detection of faces or not
  5. AutoTagging Quality (list field): control which type of quality you want to estimate

The default values can be changed on your Cumulus catalog record field settings.

1. Automatic tagging with keywords

With the tagging enabled the integration will analyze the images content and save keywords (tags) to the $Keywords tree or a text field. Both can be specified in the Cumulus Scheduler job settings. The $Keywords tags are stored within the $Keywords tree below a second level entry for every tagging model to better separate them from your own keywords.

All values are stored in Cumulus records and can be changed by a user without any restrictions. The type of a keyword (manually or created by AutoTagging) can be seen by a tooltip over every keyword within the Cumulus web client for example. If it is a keyword with one of the tagging models as a prefix then it is a keywords from AutoTagging.

Screenshot

2. Assigning colors to label fields

If you want to recognize the most dominant colors for your assets you can use this feature. The recognition uses the W3C web colors (https://en.wikipedia.org/wiki/Web_colors) and matches the nearest neighbor. The actual value will be stored in eight Cumulus label fields and will store the most dominant one in the field AutoTagColor_1 the second most dominant in AutoTagColor_2 and so on. Depending on the image you will got up to eight colors or only a few. The delivered record view set shows only the fields with a value.

3. Classification

The classification tries to recognize if there is "not-safe-for-work" content in the images and classifies it to "safe for work" or "not safe for work" if the API gets values below the probability treshold defined in the Scheduler job configuration you will get not value.

4. Face recognition

Faces can be detected if the control field "AutoTagging-Face" is activated. The result will be stored in a table field and optionally in the comments of the record to better visualize their position within the image.

5. Using own custom tagging models

To use custom tagging models you have to create a new model at your Clarifai account and train it with sample images. To use this in your Cumulus application you only have to enter the exakt model name in the record field "AutoTag Model" as a new list entry. Now you can set the value to this new entry for the records where you want to use this model for tagging.