The latest Nvidia AI system can reportedly fix or repair bad images and can even remove watermarks from photos.
The new Nvidia AI system was reportedly developed by the company’s researchers in collaboration with MIT and the Aalto University. According to reports, Nvidia’s latest innovation can automatically remove noise, grain, and watermarks visible in images.
However, what makes the artificial intelligence more impressive is that it can do all those things without the need to study any clean photo examples. Instead, all it needs are damaged photos.
“This deep learning-based approach has learned to fix photos by simply looking at examples of corrupted photos only,” Nvidia said in a statement.
“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.”
In a paper published by the researchers, they claimed that their neural network is on par with today’s “state-of-the-art methods that make use of clean examples.” The AI was trained on 50,000 images in the ImageNet validation set using NVIDIA Tesla P100 GPUs with the cuDNN-accelerated TensorFlow deep learning framework.
“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,” the team further explained.
“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 new Nvidia AI system will be presented by the researchers at the International Conference on Machine Learning happening at Stockholm in Sweden.