The Twilight actress made her directorial debut in a short film Come Swim, shown yesterday at the Sundance Film Festival and features the use of the popular neural “style transfer” technique to build its story.
The project, co-authored with an Adobe research engineer, is a great case study on the ability to use style transfer in a production. Using CUDA, GPUs on the Amazon cloud and the cuDNN-accelerated Caffe deep learning framework, they redrew key scenes in Come Swim in the style of the impressionistic painting that inspired the film.
As mentioned in the paper, GPUs played a key role in the work since they provided the computing power necessary to achieve the quality needed for the images in a reasonable amount of time.
Read more >
Actress Kristen Stewart Co-Authored AI Style Transfer Paper
Jan 20, 2017
Discuss (0)
Related resources
- GTC session: The Impact of AI on Filmmaking: Real-World Transformation
- GTC session: AI Research Careers: From Research to Reality
- GTC session: Build an AI Research Assistant With NVIDIA AI Blueprints
- NGC Containers: retail-shopping-advisor-lab-service
- SDK: TAO Toolkit
- SDK: RTX Kit
About the Authors
