On Taming Transformers for High-Resolution Image Synthesis

Impressive paper from the end of 2020 accepted at CVPR 2021! Well written. A state of the art combination of Convolutional Neural Network, Transformer and Generative Adversarial Network plus various methods.

This paper was written by a group of German researchers from the Heidelberg University. Thus, no surprise to find some German quirks (like deutsche Gründlichkeit (thoroughness) and particular attention to peculiar details) in this paper.

Thanks to the many high resolution images (too many in my opinion) included in the PDF file of this paper, it is almost 51MB large, larger than most other computer vision papers.

[2012.09841] Taming Transformers for High-Resolution Image Synthesis