Organs began as a cGAN trained to generate body organs. The next version widens the word.
An organ is the essential transformative part of a system, a machine or a body: the part that takes in matter, processes it and hands it on, ready for a new shape or a new process. A kidney, a gill, a leaf, a mycelium, a heat exchanger and a hand pump all do this. The model is meant to generate these forms, not only the ones found inside bodies.
Before a model learns anything, the data has to be made. Organs, the data follows that step: 5 808 images in five realms (animal, plant, fungal, metaphor and sci-fi) gathered from Wikimedia Commons, iNaturalist and the Wellcome Collection, cut out and centred on one neutral ground, turned into masks with a single class, and reviewed by a person. The training itself, a small network that answers a painted region with a specimen, is planned and not yet done.
The last four images show that page. The photographs on it belong to their authors and are credited there.