Text Embedding is Not All You Need: Attention Control for Text-to-Image Semantic Alignment with Text Self-Attention Maps.

Jeeyung Kim, Erfan Esmaeili, Qiang Qiu

Browse the full CVPR paper archive.