generated at
ECLIPSE
https://huggingface.co/papers/2312.04655ECLIPSE: A Resource-Efficient Text-to-Image Prior for Image Generations
https://arxiv.org/abs/2312.04655ECLIPSE: A Resource-Efficient Text-to-Image Prior for Image Generations
https://eclipse-t2i.vercel.app/Project
Demohttps://huggingface.co/spaces/ECLIPSE-Community/ECLIPSE-Kandinsky-v2.2ECLIPSE-Community/ECLIPSE-Kandinsky-v2.2
Codehttps://github.com/eclipse-t2i/eclipse-inferenceeclipse-t2i/eclipse-inference
Modelhttps://huggingface.co/ECLIPSE-Community/ECLIPSE_KandinskyV22_PriorECLIPSE-Community/ECLIPSE_KandinskyV22_Prior

Kandinsky 2.1よりもパラメータ数が少なくSDXLレベルの性能
KarloとWürstchen

データセット：MS COCO/CC3M/CC12M/HighRes
ベンチマーク：FID/T2I-CompBench
Stable Diffusion 2を0から作るコストは16万ドルを更新?