Tag: #ViT

← All posts

• Self-Supervised Learning 2: MAE — Why Covering Three-Quarters of an Image and Guessing It Works So Well

  Just as fill-in-the-blank in text (BERT) powered up language models, covering up pieces of an image and guessing them helps models learn great features. We explain why MAE masks 75% instead of BERT's 15% (images have far

  2026-10-17 · 11 min read #비전 모델#자기지도 학습#MAE#BEiT#마스킹

• Vision Transformer: Reading an Image Like a Sentence of 16×16 Patches, and Why It Needs So Much Data

  We examine the principles of ViT (2021), which reads images using the Transformer originally designed for reading sentences. It cuts an image into 16×16 patches, lays them out like words, and uses self-attention to exami

  2026-10-17 · 13 min read #비전 모델#ViT#트랜스포머#셀프 어텐션#위치 임베딩