1 paper · 2 filters
Sunung Mun, Sunghyun Cho, Jungseul Ok
Recent text-to-image (T2I) generators can synthesize realistic images, but still struggle with compositional prompts involving multiple objects, counts, attributes, and relations.…