ArtSC

What if your artistic medium could expand your network ? ArtSC (pronounced 'artsy') uses AI to predict the genre of your creative work and shows creators with similar pieces who you can connect with.

Comment

Inspiration

As friends who use art as a creative pastime, we wondered if there is a way to connect artists based on their shared artistic style. A web application where an artist can find other artists who work with the same medium could foster collaboration and the formation of supportive communities.

What it does

  1. Provides artists with a web and mobile social media platform to share their work
  2. Uses Artificial Neural Networks to predict the genre of a shared piece of art (sketch, portrait, etc...)
  3. Displays profiles of other artists that have also shared works in the same genre as the poster.

How we built it

  1. The web application was built using ReactJS to offer essential functionalities like Login/Signup, Posting images and displaying users with similar artwork. The layout and design was done using React Bootstrap.
  2. The mobile application was developed on iOS.
  3. A Django backend offers API end points to the web and mobile apps
  4. The Neural Network model was trained using a dataset from here. Images from the 15 most prevalent classes were included in the final dataset. The Inception V3 was used as the backbone to perform transfer learning by fine-tuning performance on this dataset.

Challenges we ran into

  1. Gathering data to train the model was a significant challenge as there was no previously existing corpus
  2. Several issues regarding layout and placement of buttons/grids in React

Accomplishments that we're proud of

  1. Made a web and mobile application in such a short time.

What we learned

  1. A seemingly straightforward idea takes a lot of work to materialize.
  2. Working with teammates from diverse backgrounds puts many useful perspectives on the table and allows for the creation of better end-products.

What's next for ArtSC

  1. Enhancing the chat functionality to support real-time conversation
  2. Extend the number of image classes the neural network can identify.
Share this project:

Updates