Photo Credit: Nvidia
In a new advancement, an AI system developed by a team of researchers that allows you to erase all the noise from a photo. Be it pixelation, watermarks, or even text. This new AI image restoration technique enhances grainy images as well, and this new method has been found by folks at Nvidia, MIT, and Aalto University. This AI system is powered by deep learning neural network that has been trained with the help of 50,000 photos, to get an idea of what a noise-free end result should ideally look like. The system "learned to fix photos by simply looking at examples of corrupted photos only."
The research report submitted by the team suggests that the AI engine can restore images better than a professional photo restorer. The researchers included in this project are Jaako Lehtinen, Jacob Munkberg, Jon Hasselgren, Samuli Laine, Tero Karras, Miika Maittala, and Timo Aila. They used Nvidia Tesla P100 GPUs with the cuDNN-accelerated TensorFlow deep learning framework, and trained their system on 50,000 images in the ImageNet validation set. They never showed the system what a noise-free image looks like, and even without a before and after picture training, this AI can remove artefacts, noise, grain, and automatically enhance photos. While this doesn't require a clean image for learning purposes, it does need to see the source image twice before making the changes.
"Without ever being shown what a noise-free image looks like, this AI can remove artifacts, noise, grain, and automatically enhance your photos," a blog post reads.
"There are several real-world situations where obtaining clean training data is difficult: low-light photography (e.g., astronomical imaging), physically-based rendering, and magnetic resonance imaging. 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. Of course, there is no free lunch - we cannot learn to pick up features that are not there in the input data - but this applies equally to training with clean targets," the team writes in its paper.
The paper submitted by the researchers also notes that this AI system has many real-world applications like it can be used to enhance MRI images, perhaps paving the way to improved medical imaging.