Monet

Monet is an experiment in evolving image filters with a genetic algorithm. How it works, end to end:

• Training images — JPGs dropped into an upload folder are analyzed (color histogram), assigned a hash, moved into a training set.
• Population of filters — each "individual" is a set of genes, where genes are parameter presets for CamanJS image filters (brightness, contrast, saturation, etc.). The first generation is random.
• Rendering — every training image is re-rendered through every individual's filter preset.
• Voting — the browser client shows a photo in its various filtered versions; users vote for the one that looks best.
• Evolution — over enough votes, the fittest filter presets are bred into the next generation, so the filter parameters gradually improve.

What's next for Monet

project abandoned :(

Built With

• html5
• javascript
• mongodb