ICML2025
Event-Customized Image Generation
Zhen Wang, Yilei Jiang, Dong Zheng, Jun Xiao, Long Chen
摘要
<V> sleeping <V> in a doghouse a girl wearing a hat and a scarf <V> Spiderman <V> a panda <V> <V> monkey <V> monkey cat <V> cat <V> girl , hat scarf <V1> <V2> ① <V1> ② <V2> ③ cat (a) The subject customization (b) The action and interaction customization (c) The event customization Figure 1: Customized Image Generation. (a) Generating customized images with given subjects in new contexts. (b) Generating customized images with co-existing basic action or interaction in given images. (c) Generating customized images for complex events with various target entities. Different colors and numbers show associations between reference entities and corresponding target prompts.