Virginia-based artist Pindar Van Arman and his two kids built a deep learning painting robot that makes all artistic decisions without human intervention.
The robot named CloudPainter has painted over 1,000 canvases in the last ten years with a combination of AI and human direction.
“I started experimenting with various artificial intelligence algorithms and tried each of them to make the robot more independent,” mentioned Van Arman in a video explaining the technology behind the creativity.
Besides scheduling the photoshoot, CloudPainter makes its own creative decisions. The robot snaps photos of the human subject and creates an original abstract composition from its favorite by using a variety of aesthetic decisions and facial detection – measuring symmetry, proportion and the direction the subject is looking – it then executes each brushstroke until it has calculated it had done the best it could to render its original abstract composition… similar to human artist.
Their most recent robot incorporates the popular “style transfer” technique. Using a GTX 1080 GPU and cuDNN with the TensorFlow deep learning framework, the robot’s original composition is reimagined in the style of a reference piece of art – either of one that it made itself, a famous painting, or a painting by one of Van Arman’s kids.
CloundPainter AI was recently featured in the GPU Technology Conference opening keynote video “I Am AI”.
Learn more about CloudPainter >
Related resources
- DLI course: Building Real-Time Video AI Applications
- GTC session: Create Your Artistic Portrait with Multimodal Generative AI and NVIDIA Technologies
- GTC session: Autonomous Mobility Robotics is the Key Driver Accelerating the Era of AI (Presented by MSI)
- GTC session: AI and the Data Universe: Refik Anadol’s Artwork for Global Impact
- NGC Containers: quickstart-rapidsai
- SDK: Isaac Manipulator