[2010.11929] An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
5
公开标注数
5
参与人数
2026-07-17 10:04:36
首次 Whisper
本页的公开 Whisper
划选高亮2026-07-17 13:25:36
原文高亮摘录
“attention is either applied in conjunctio”
Whisper 随想笔记
I always thought CNN+attention was the way, guess this flips it.
划选高亮2026-07-17 13:16:36
原文高亮摘录
“attention is either applied in conjunctio”
Whisper 随想笔记
So they're saying attention doesn't need convnets to lean on? That's a bold claim.
划选高亮2026-07-17 10:22:36
原文高亮摘录
“An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale”
Whisper 随想笔记
I tried this once, my GPU cried. Paper's good though.
划选高亮2026-07-17 10:13:36
原文高亮摘录
“An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale”
Whisper 随想笔记
16x16 words, more like 256 patches, but okay, catchy title.
划选高亮2026-07-17 10:04:36
原文高亮摘录
“An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale”
Whisper 随想笔记
So transformers finally came for images, huh? wild times.
分享本页 Whisper
短链接
https://domwhisper.com/s/4aba4475c526嵌入代码
<iframe src="https://domwhisper.com/embed/4aba4475c526" width="100%" height="480" style="border:0;border-radius:16px" loading="lazy"></iframe>