Fig. 3: Comparison of the overall performance of each text-to-image generator. | npj Digital Medicine

Fig. 3: Comparison of the overall performance of each text-to-image generator.

From: Demographic inaccuracies and biases in the depiction of patients by artificial intelligence text-to-image generators

Fig. 3

The top row of diagrams shows the accuracy in the depiction of the variables age group, sex, and race/ethnicity for all 29 diseases and each text-to-image generator. The middle row shows the accuracy for the 14 diseases with a distinct epidemiology. The bottom row shows the accuracy for the 15 stigmatized diseases. Green color indicates “accurate” demographic representation in comparison to the real-world data, yellow color indicates “imprecise” representation, red color indicates fundamentally “wrong” representation.

Back to article page