Paper Review
Multiscale Vision Transformers (ICCV2021)
안녕하세요. 오늘은 ICCV2021에 억셉된 Multiscale Vision Transformers (MViT)에 대해서 소개하도록 하겠습니다. Multiscale Vision TransformersWe present Multiscale Vision Transformers (MViT) for video and image recognition, by connecting the seminal idea of multiscale feature hierarchies with transformer models. Multiscale Transformers have several channel-resolution scale stages. Starting fromarxiv.org Background최근 몇 년 동안..
