Just wrote this article introducing the main ideas on why Transformers can be so powerful for Computer Vision applications. Hope you find it insightful, any feedback is welcome :)