201217 Taming Transformers for High-Resolution Image Synthesis.md

https://arxiv.org/abs/2012.09841

Taming Transformers for High-Resolution Image Synthesis (Patrick Esser, Robin Rombach, Björn Ommer)

vq-vae로 이미지 크기를 줄인 다음 트랜스포머로 prior를 모델링. 결과 퀄리티가 인상적. https://compvis.github.io/taming-transformers/

#discrete_vae #generative_model #autoregressive_model #transformer