Semantically consistent cross-domain image translation facilitates the generation of training data by transferring labels across different domains. However, existing generative models struggle with object-level accuracy when translating images between domains...