ART-ificial

It is an AI that generates new art works using a generative neural network.

Comment

Inspiration

The inspiration for this project came from a YouTube video about AI generated music. I knew that with my background in computer vision that generating new art would be a good place for me to start.

What it does

ART-ificial takes a data-set of ~240,000 images and uses that as a base to build a neural network that is capable of creating new works of art.

How I built it

I found examples of other generative neural networks and used them as an example of how to build mine. The entire project lives on Google Cloud due to the cheap compute resources. The images generated by the AI are served up by a flask api that is connected to the Storage Bucket that all of the data for the project is stored in.

Challenges I ran into

Google Cloud GPU quotas took a long time to get increased, leaving me to only use CPU's to train my models. This resulted in a massive delay in development.

Accomplishments that I'm proud of

I am happy that it doesn't crash most of the time for one thing. I am also happy to know that I learned how to use machine learning to generate new content.

What I learned

How to use machine learning to generate new content.

What's next for ART-ificial

I am considering trying to improve the quality of the images produced by the network to the point that they could be used as designs for things like stickers.

Built With

Share this project:

Updates