Inspiration

Visual art has always been a medium of expression for all irrespective of social and economic lines that have been drawn in human minds across the ages. There is something for everyone to relate to, emulate, express, and enjoy. But, the enjoyment and commentary of popular and celebrated visual art were reserved for the wealthy for a long time. Cultural and stylistic diversity that was non-eurocentric is rarely appreciated in visual artistic choices. The advent of the internet and digital art opened this space to a much wider set of artists and audience, quickly breaking down the barriers of a traditionally euro-centric art world without confining the viewer to a particular gallery, style, artist, or location. DeepART is conceived as a step to democratize this further and take an experiential approach towards identifying, learning, and appreciating art. Socially and economically challenged people in developing nations do not have exposure to the therapeutic experience of art education and practice. DeepART aims at Opening up this space for the next billion users. The DeepART Gallery functions as a repository of artists and artwork in place of a traditional art gallery. While art is pivoting and branching out from traditional painting and moving more towards highly digital formats, the deepART gallery can be a one-stop solution for art enthusiasts and to view, analyze and experience virtual art and traditional art both in a single immersive digital space.

What it does

We have created an algorithm that identifies the artist when provided with a painting, with the state-of-the-art precession by using a deep convolutional neural network. By using this network we can identify the artists with an approximate accuracy of 99% on the training set and ~87% on the cross-validation set. When a user enters the image URL the model tries to identify the relevant artist with the predicted accuracy. We have also added an interactive augmented reality-based Art museum with the help of C4d and Echo-AR where users can have a dynamic view of the Art environment.

How we built it

We used a deep learning-based CNN model to recognize objects in a photo, then we uploaded the dataset into dropbox for easy access, later we created a 3d model of each 11 artists to display it in Augmented reality for having user interactivity.

Challenges we ran into

  • Understanding the architecture of CNN to identify the artist based on their paintings.
  • Working to improve and implementing our 3D model on cinema 4d.
  • Cloud connection requirements and deployment

Accomplishments that we're proud of

Before this hackathon, our team was not very familiar with Deep learning frameworks and building 3d models in C4d. While developing the project along the way we learned more about combining the artwork with different technology. We effectively manage to work in a multi-disciplinary team with members from different domains.

What we learned

Our team learned a lot about deep learning and making 3d models, EchoAR technology, and brain-stormed on different ideas to combine the application of art with deep learning.

What's next for DeepART

  • Currently, when an image is uploaded or a url of an image is provided by the user, deepart recognizes the style and stroke - and identifies the artist. In the future, we are planning to expand this to include non-euro-centric artists and digital artists in non-traditional styles - such as 3D, VR art, and further into contemporary dance
  • User can input images to train models in the future for new and emerging artists as well, thus keeping deepART inclusive and growing.
  • Currently, our gallery has space for viewing a few paintings from the 11 master artists on which the data set has been trained. In the future, this can be expanded to hold information about the complete set of works of acclaimed artists from across the world, and also digital and VR artists whose art can be experienced in VR.
  • The gallery can also be a digital space for the experience of AR NFT Art
  • A digital space like a seminar room can function to open up courses and conversations about the practice of visual and performance art where anyone with an internet connection can join into the conversations, observe and learn.

Built With

Share this project:

Updates