Link to Documentation in PDF : stemai.github.io/
The challange is to make beatiful and engaging activities that helps children learn.. period. We decided to stick to web based interface hence we are using TensorFlow.js and JavaScript implementation of Magenta.
Communication is the key part of how humans convey ideas to one another. From a young age children develop the ability to depict objects, and argubly even emotions. So our basic idea is to create something that can help them feel more connected to their surroundings which is we believe-- is beginig of education.
A paper called, "A Neural Representation of Sketch Drawings", presents a generative RNN or what we commonly called G-RNN is capable of producing sketches of common objects. We tried to implement the idea and trained our model on the dataset of some common hand-drawn sketches.
Here in this work, we investigate a lower-dimentional vector-based representation inspired by how people draw. This model, sketch-rnn, is based on sequence-to-sequence autoencoder framework.