Learning Disentangled Identifiers for Action-Customized Text-to-Image Generation

<A> "A boy <A>" "Spiderman <A>" "Messi <A>" "A gorilla <A>" "A bear <A>" "A panda <A>" Sample Images "An old man <A>" "Batman <A>" <A> "Barack Obama <A>" "A monkey <A>" "A polar bear <A>" "A cat <A>" Sample Images <A> "A woman <A>" "David Beckham <A>" "Michael Jackson <A>" "A dog <A>" "A fox <A>" "A cheetah <A>" Sample Images Figure 1 . Action customization results of our ADI method. By inverting representative action-related features, the learned identifiers "<A>" can be paired with a variety of characters and animals to contribute to the generation of accurate, diverse and high-quality images.