Rethinking Spatial Dimensions of Vision Transformers
Byeongho Heo1 Sangdoo Yun1 Dongyoon Han1 Sanghyuk Chun1 Junsuk Choe2 * Seong Joon Oh1
             1           2
               NAVER AI Lab      Department of Computer Science and Engineering, Sogang University

               Abstract                   eration on ImageNet [8]. As a result, a new direction of net-
                                      work architectures based on self-attention mechanism, not
  Vision Transformer (ViT) extend ...