Using Google’s Artificial Intelligence to tag images
Modern computer vision systems can, potentially, save a lot of time and effort when it comes to tagging generic content. For example, stock images and pictures of common objects can be tagged with very good results. Third Light integrates with Google's Cloud Vision system to provide this facility as part of Chorus.
In the example, below, specific metadata fields have been created inside a Chorus space. These fields are separate from any other metadata on the site, so that it's easy to distinguish machine tagging from manual (human) tagging.
The image has a number of high-quality metadata tags, provided by Google Vision, which make the file much more useful in the Chorus search engine. The tags were obtained by sending the preview image to Google, which returns the metadata suggestions a few seconds later. In Chorus, this can happen automatically via the Third Light API.
Other than the time saving, one of the advantages of using a computer vision system to tag images is that obvious themes and connections between files are then easy to use and navigate through Chorus. At the top of each folder in Chorus, we display metadata facets which offer a simple, clickable way to refine the view by metadata "facets". And, since the Google Vision keywords are stored separately, it's easy to filter them out when required.
The Google Vision module in Chorus supports label detection and landmark detection.