Convolution is not dead: A ConvNet for the 2020s- Machine Learning, Image Classification Paper React

preview_player
Показать описание


Pros and Cons Convolution vs Transformers: 0:50
Why it worked (Large Kernal Sizes, fewer activations): 1:49
My thoughts when to use convolution vs transformers: 6:53

10 Jan 2022 · Zhuang Liu, Hanzi Mao, Chao-yuan Wu, Christoph Feichtenhofer, Trevor Darrell, Saining Xie · Facebook Research, UC Berkley

Рекомендации по теме
Комментарии
Автор

Thank you for making this video. Of course it is not dead because training transformer takes time.

jonathansum