Automatic Image Tagging

Imagga looks at an image and returns a list of tags describing what's in it, such as objects, scenes, or colors, without a person labeling it by hand. A new image uploaded to a catalog or media library is what a Beam agent requests tags for, reading back the list Imagga returns and writing the tags onto the image's record so it can be searched or filtered later. Tags that come back with low confidence, or nothing usable at all, leave the record untagged, with a person notified to add tags manually instead of an unreliable guess getting filed.

Visual Similarity Search

Imagga can compare an image against a library of others and return the ones that look visually similar, useful for catching duplicate or related photos. A new image uploaded to a catalog that shouldn't contain duplicates is what prompts a Beam agent to run this search, reading back any close matches Imagga finds and flagging the record if a near-duplicate already exists rather than adding it a second time. Images that are similar but not clearly the same photo are left for a person to compare side by side and decide.

Explicit Content Moderation

Imagga checks an image for explicit or otherwise flagged content and returns a score for how likely it is to violate a policy. New uploads go through this check as they come in, with a Beam agent reading the score Imagga returns and applying the account's approved threshold to either publish the image or hold it back automatically. Anything scoring in a gray area, rather than a clear pass or a clear violation, is routed to a moderator to review by eye instead of being published or removed on the agent's own call.