Inspiration
I thought about building this after seeing simple sketches. I thought, AI has the ability to mimic the human brain, so why not train it to be able to autocomplete images. Using unsupervised learning, and a generative RNN Network, we were able to do this.
What it does
It autocompletes any sketch you start to do pretty well, and easily I might add.
How we built it
We started as a JS backend and CDNs for the packages. We then built this around a canvas object in DOM HTML. Then we made CSS and hosted it on netlify.
Challenges we ran into
It was very difficult to figure out how to complete this feat. Thinking about it, what we did is pretty much impossible. How we did it is by taking the partially drawn image they drew, then the AI will get its model, and compare and see which parts are missing. Then it fills it ins. This is how it's currently doing it. But the cool thing about AI is that they have the ability to alter their own code, therefore it's always constantly learning new ways to sketch more accurately.
Accomplishments that we're proud of
Well, first off, completing this. This was no simple task to do, and I wanted to do it using only CDN elements, which makes this very hard to do.
What we learned
How to build AI using CDN elements
What's next for Auto Sketch AI
plan on training it on more complicated sketches to complete professional sketches instead of doodles.
Log in or sign up for Devpost to join the conversation.