× Projects Active and Future Projects

Nostalgia Box

Machine-learning based digital memory soup

Nostalgia box is an assortment of images, blended together into a soup of memories via the combined powers of machine learning and simple animation. It is an extension on the paper A Neural Algorithm of Artistic Style by Leon A. Gatys, Alexander S. Ecker, and Matthias Bethge. Nostalgia Box is a continually shifting soup of its user's memories. Images, curated by the user for their nostalgic emotional impact, are dreamed over one another using a neural style algorithm. The resulting images are then overlaid into a video which never fully solidifies into any one image, but instead is always several at once. The user themselves are able to recognize what they see, but after losing track of what images they've entered, should lose their certainty in their identifications. A stranger should see only a haze of vague shapes which occasionally contains the suggestion of a face. This is machine learning, stripped of the elements of spam. You controlled what information went into it, you can understand what is being created with it, and no one is trying to sell you anything. It contains 58 images created using Deep Dream Generator. It uses a Javascript animation to fade through these images.

Visit the box
Media Lab Project Page
Download the code on Github
Check out our art collective