1 paper
Ryota Okumura, Kaede Shiohara, Toshihiko Yamasaki
Recent text-to-image models, such as Stable Diffusion, have achieved impressive visual quality, yet they often suffer from geometric inconsistencies that undermine the structural r…