NVIDIA demonstrate latest denoising AI that works only by looking at other grainy photos

NVIDIA has released information about their latest AI that can remove noise and artefacts from photos by being trained on examples of other similarly corrupted images. According to NVIDIA “Recent deep learning work in the field has focused on training a neural network to restore images by showing example pairs of noisy and clean images. The AI then learns how to make up the difference. This method differs because it only requires two input images with the noise or grain.”

It is possible to learn to restore signals without ever observing clean ones, at performance sometimes exceeding training using clean exemplars, [The neural network] is on par with state-of-the-art methods that make use of clean examples — using precisely the same training methodology, and often without appreciable drawbacks in training time or performance.

The work was created in a collaboration between NVIDIA, Aalto University, and MIT, and will be presented on July 12 at the International Conference on Machine Learning in Stockholm, Sweden this week. You can read more on NVIDIA’s developer website and see a demonstration on YouTube.