This is the text version of the interactive 3D documentary. 한국어
Contents · Chapter 3 One photo
Put one of your photos into the network
Key ideaThe network tries to rebuild the whole from a piece
Add a photo with the button below. It is turned into a 24×24 grid of light and dark dots and stored in the network along with 5 pictures. Then only its top-left quarter is shown, and you can see whether the network fills in the rest. Your photo is processed only on this device and is never sent anywhere.
Why?If the photo fits the board well, it comes back almost exactly. If it looks like another picture, it may blend or turn into that picture. Without a photo, a smiling face is used.
Go furtherPhotos with clear light and dark areas, like faces, work best. A 24×24 grid is tiny, far rougher than the scenes a person remembers.
View in 3D →