CrossPrism AI Training for Photo and Video Keywording/Labeling

CrossPrism offers several advanced methods of training depending on your needs.

Multiple Choice (Multiclass)

Simply drag and drop photos to associated labels. As more sample photos are manually associated across different labels, CrossPrism automatically learns the particular traits that belong or don't belong to a label.

Dual Choice (Binary)

This specialized case is useful for filtering is/is not (pass/fail, like/don't like, etc.) type of operations. It uses only one label for a positive id; a negative id is indicated by the absence of the label.

Descriptive (Zero Shot)

This state of the art technique uses the latest pre-trained "transformer" types of models: They already have an understanding of photos and language. The only necessary training is to specify which labels to focus on.


Best for finding burst sequences or duplicates, cluster training groups similar photos together without the need to provide examples.