media · Picked by us
A camera that can only see what it already knows
Memo Akten points a camera at the world and rebuilds it through neural networks trained on everyday images — so the machine sees everything through the filter of its own experience.
SIGGRAPH-published account of the models and the installation
Official project page with the video works; media stays on the artist’s site
What it does
A series of videos and a real-time installation where a trained model reconstructs whatever the camera points at, using only what it learned before.
Who it helps
People interested in machine perception, generative art and the limits of models.
How it was made
- Models
- pix2pix · image-to-image GAN — creator-reported
- Tools
- TensorFlow · Real-time image-to-image inference
- Akten trained image-to-image models on selected image collections so the network builds its own visual priors.
- Live camera frames are preprocessed and fed through the model in real time, with temporal smoothing to steady the output.
- The public code and a SIGGRAPH paper document collection, training, inference and the exhibition history.
What the person did
Concept, training choices, curation and exhibition iterations are the artist’s work; the piece is authored rather than generated.
What we know
Creator-reported: First-person project page, public code and a SIGGRAPH paper; the GitHub profile resolves to the same artist.
Access checked — 2026-09-26 · Opened the project page, the implementation repository, the paper and the creator profile on 2026-09-26.
Can you use it?
The artwork and its videos stay on memo.tv and its official Vimeo embeds; we link only. The demo code is MIT. The cover is an original Made With Models illustration, not a frame from the work.
About this project
Picked by Made With Models. Added 2026-09-26. Updated 2026-09-26.