PEIT: Bridging the Modality Gap with Pre-trained Models for End-to-End Image Translation

37806

Knowledge Triples