Draw to Art is a new experience that uses machine learning to match your doodles to drawings, paintings and sculptures from museums around the world. We created a series of interactive easels so everyone can use Draw to Art in a tangible way in events and cultural spaces.

With Draw to Art, everyone can now use their own creativity to discover artworks from the Google Arts & Culture collection in a whole new way.

Using machine learning, we trained a deep neural network to recognize visual features in doodles, like shapes, lines and perspectives. We also trained it to recognize the same features in paintings, sketches and sculptures from the collection so Draw to Art can associate doodles to similar works of art.

The best matches are then presented to the user who can help teach the neural network by rating the matches to help improve the results for everyone.

Draw to Art was designed to be very easily usable by everyone with little explanations - Draw something on the left of the canvas and visually similar artworks are presented on the right.

We spent the first month of the project testing out different machine learning models and prototyping different user interfaces - we then finalised the digital design and worked with a product designer to create the interactive easels.

Draw to Art was unveiled to museum partners & influencers on the 5th of March in the Google Cultural Institute lab in Paris where it’s now permanently installed and is used daily by visitors who ranges from international politicians to school childrens. Visitors spend 10 min+ on average using the easels.

Google Arts & Culture now has an easily scalable installation that everyone can use to discover artworks in a whole new way using machine learning.

It is already scheduled to be installed in 6+ events and exhibits in Europe and the US in May and April and will be installed on an ongoing basis in cultural spaces and events.