Creators can benefit from this world-first AI method, which works by simply looking at examples of grainy photos instead of clean imagery.
Even the most talented creator dreads the prospect of cleaning up a grainy photo, with the denoising process always needing the input of ‘cleaner’ images in order to facilitate the process. Changing all that though is work developed by tech company Nvidia with researchers from Aalto University and MIT, whose deep-learning method can fix photos simply by looking at examples of corrupted photos only.
In work that will be unveiled at the International Conference on Machine Learning in Sweden this week, the team’s method differs as it only requires two input images with the noise or grain, from which a denoising AI can remove artifacts and automatically enhance your photos.
Using Nvidia Tesla P100 graphics cards with the cuDNN-accelerated TensorFlow deep learning framework, the team trained their system on 50,000 images in the ImageNet validation set, validating the neural network on three different datasets.
While this will be of great interest to the design community, more practical applications of the AI could see it being used to enhance MRI images (below) and astronomical snapshots.
“There are several real-world situations where obtaining clean training data is difficult,” the research team explain in their paper on the project, “(and) our proof-of-concept demonstrations point the way to significant potential benefits in these applications by removing the need for potentially strenuous collection of clean data.”
The team’s paper will be presented this Thursday at the ICML conference – though there’s no news on where this will be licensed to applications developers so it could be put to good use by photographers, designers and other creators. In the meantime, learn how to clean photos with this Photoshop tutorial on how to remove noise and sharpen an image.