Heya in English means "She's". Heya the passionate, hardworking, daughter, sister, wife, and/or mom. The artwork is inspired by the concealed & unknown/unrevealed energy and power of Heya.

and, two of the numerous Heyas in the whole world have came up with this collection of artworks using AI techniques at AIArtathon :)

Team's Bio:

The artists behind this artwork Reema Albaz and Aeshah Alabdullatif, share a strong interest in digital art creation and AI. Albaz is an experienced digital designer and Alabdullatif is a data scientist, ex-web developer and designer. As a coder and developer, Alabdullatif has a lot of experience in working with math and statistics, but she also has a big interest in sketching and coloring. In her work Alabdullatif loves to experience with color palettes using coloring apps and Faber-Castle color pencils and she often uses digital visualization tools to interpret her ideas into simple, informative, and eye catching pieces of art. Albaz has also worked extensively with sketching and she often uses Adobe's creative digital tools such as Adobe Draw for iPad to sketch on the go.

The main reason for the team's interest in the field of artificial intelligence is a shared eagerness to express their thoughts and minds through an immediate and self-explanatory work of art, that does not require any prior knowledge from the audience.

What it does

The video is a documentation of the actual artwork, which is interactive. The actual artwork consists of a screen and a camera. Women are invited to stand in front of the screen. When a woman stands in front of the screen, the camera takes a picture of her, which is sent to an algorithm. The algorithm detects the person's specific facial tones and hair color and based on these characteristics a sketched image of a woman with similar facial tones and hair color is shown on the screen. Each new sketched image has a new random calligraphy textures pattern in the background, which is generated by the previously mentioned StyleGAN model.

How we built it

The dataset of sketched images of different women have been digitally hand drawn by the artist team. The artists have created 4 different styles of women with different hair length and hair styles. They have used a unified color coded to make it easy for the AI and algorithm. The drawing style and the backgrounds are inspired by the mysterious, arcane and powerful energy of Heya. The artists believe this work can reflect the ability to match or shorten the realism distance between a photographed selfie and the digital, sketched ones.

Challenges we ran into

It is our first trial to use the AI with images, thus we had some difficulties of setting proper development environment and hardware capabilities been an issue. Other challenge is the detection of the hair features, since there is a low accuracy in the literature, but we could come up with a good hair features detection algorithm.

Accomplishments that we're proud of

The piece of art that our code is generating at every run. The knowledge we made to collect during this period of time. The algorithms built to generate what we can call an impressive artwork.

What we learned

  • How to construct an art concept
  • How to apply StyleGANs & BigGAN and the neural style transfer methods
  • How to apply some computer vision approaches to preprocess the image
  • The possibilities to overwrite a good detection algorithm

Making this artwork has been an exciting journey for both artists and made them realize the magical potential in making artworks using artificial intelligence.

What's next for 11_Heya

The ideas behind the project Heya are scalable to many choices and add-ons which can make a good profitable artistic start-up.

Built With

  • biggan
  • cv
  • neuralstyletransfere
  • python
  • runwayml
  • stylegan2
Share this project: