论文简介

原论文：Video Swin Transformer¹

论文地址：https://arxiv.org/abs/2106.13230

以下仅为作者阅读论文时的记录，学识浅薄，如有错误，欢迎指正。

论文内容

摘要

• The vision community is witnessing a modeling shift from CNNs to Transformers, where pure Transformer architectures have attained top accuracy on the major video recognition benchmarks.
  计算机视觉领域正在见证着模型从CNN到Transformer的变迁，纯Transformer结构已经在主要的视频识别基准上达到了最高准确率。

• These video models are all built on Transformer layers that globally connect patches across the spatial and temporal dimensio