Training neural networks to perform tag-based font recognition utilizing font classification

A non-transitory computer-readable medium storing instructions that, when executed by at least one processor, cause the at least one processor to perform operations comprising:

• receiving font training images, corresponding font tags, and corresponding font classifications;
• training a font tag recognition neural network to determine initial tag-based font probabilities for fonts using the font training images and the corresponding font tags;
• training a font classification neural network to classify the fonts using the font training images and the corresponding font classifications;
• and
• learning updated parameters for the font tag recognition neural network that cause the font tag recognition neural network to generate enhanced tag-based font probabilities based on font classifications generated by the font classification neural network.

In a digital medium environment for creating or editing electronic documents, a computer-implemented method of recognizing fonts, comprising:

• training a font tag recognition neural network to determine initial tag-based font probabilities for each font of a plurality of fonts using font training images and corresponding font tags;
• training a font classification neural network to classify a plurality of fonts using the font training images and corresponding font classifications;
• performing a step for jointly training the font tag recognition neural network with an implicit font classification attention model within the font classification neural network to generate font classification attention maps;
• and
• generating, by the trained font tag recognition neural network, a first enhanced tag-based font probability for a first font of the plurality of fonts based on a first font classification attention map.