Spatial Asset Recommender – Introduction
Before diving into the details of the Spatial Asset Recommender, it is important to get on the same page about names and the basic functionality. I try to keep this part short and easy to understand.
From a user’s point of view, the core of the Spatial Asset Recommender is a Content Browser-like window that updates automatically when the level viewport changes. It tries to recommend assets that match the current viewport location in the level by looking at the assets that are already placed.
To find matching assets, it compares collections of weighted tags that are assigned to each asset.
The other core part is about how to get these collections of weighted tags for each asset. This process should be as automated as possible because nobody likes to manually tag gigantic lists of assets. For that purpose, classifiers have the job to do that based on their specialization. While one classifier maybe takes the tags from a descriptive text of a Blueprint property, another classifier might take a look at the visual appearance in the thumbnail or actually listen to the audio of a sound file.
While the Spatial Asset Recommender already provides a few classifiers, development teams can write their own classifiers for their specific purposes.