SESKStuff Everyone Should Know
SESK: Stuff Everyone Should Know · sesk.info/ai-vision-transformer-patches

How does a Vision Transformer (ViT) feed an image into a Transformer?

As a sequence of image patches, treated like words.

Start playing free

  1. One pixel at a time, as text
  2. As a sequence of small image patches ✔
  3. Through a convolutional network first

Why: ViT cuts the image into a grid of small squares, turns each into a vector, and feeds them in order like tokens. The 2020 paper showed this works well without any convolutions.

Huh, didn't know that: The paper's title is An Image is Worth 16x16 Words.

The dad joke
A Vision Transformer looks at a picture like I do a puzzle. One square at a time, and still confused.
Like this card?
Play the game

References

Answer: As a sequence of small image patches. ViT cuts the image into a grid of small squares, turns each into a vector, and feeds them in order like tokens. The 2020 paper showed this works well without any convolutions.

  1. Dosovitskiy et al., An Image is Worth 16x16 Words (arXiv 2010.11929) arxiv.org/abs/2010.11929
    a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks
    Checked 2026-10-10.

Think this answer is wrong? Tap "Challenge this answer" on the card in the game and tell us why.

Test yourself in the game

Artificial IntelligenceWhat was OpenAI's CLIP trained to predict on 400 million internet images?Artificial IntelligenceHow do AI, machine learning and deep learning fit together?Artificial IntelligenceWhat did Alan Turing call his 1950 test of machine intelligence?

That's stuff everyone should know.