Removing grain is a key feature photographers use in programs like Lightroom or Capture One. The ability to reduce noise and grain still has a long way to go, but research from NVIDIA shows just how far deep learning based technology is coming.
NVIDIA AI removes grain and watermarks from photos
In the beginning of the video NVIDIA explains that their technology can learn image restoration without clean data and can enhance grainy photos through AI. Without ever seeing what a noise-free image looks like, this AI can remove artifacts, noise, grain, and automatically enhance photos.
The training network can accelerate de-noising to minutes instead of days using data generated on the fly. The team of researchers from NVIDIA, Aalto, and MIT used NVIDIA Tesla P100 GPUs with a deep learning framework to train their AI system using 50,000 sample images.
The work was developed by researchers from NVIDIA, Aalto University, and MIT, and is being presented at the International Conference on Machine Learning in Stockholm, Sweden this week.
Potential Use Cases for NVIDIA noise reduction AI
There are a number of potential ways this AI can be used to aid in various industries.
Magnetic Resonance Imaging
Low-light Photography (various industries)
Read more about the research at NVIDIA on their developer blog.
The team will present their work in an oral presentation and a poster at the ICML conference. If you’re in Stockholm, Sweden for the conference, you can meet the team on Thursday, July 12 at the Supervised Learning oral session (2:20 pm) and 6:15 pm poster session.