Scientists from the Swiss Federal Institute of Technology, ETH Zurich, have been using deep learning neural networks to turn smartphone photos into ‘DSLR-quality’ pictures.
The researchers have used the technology that underpins the smart attributes of artificial intelligence systems, to turn uploaded smartphone photos into shots that ape those captured with more expensive and larger DSLR cameras.
The sample photos are impressive, adding more vibrancy and exposure than the phone photos, though there’s a slight greenish hue to some of the pictures which indicate the process in not flawless yet.
Through using a deep learning neural network. which in simple terms is an artificial take on how our brains use a web of neurons and synapses to effectively pick apart what we are seeing and make sense of our surroundings, the researchers trained the system through using sets of the same photos captured on both DSLRs and smartphones.
Through a repeated process of filtering these photos through the neural network, the smart system was able to learn how to improve smartphone pics to get close to DSLR equivalents. The system is at a level where it can now take any two sets of photos from different cameras to figure out on how to stitch together higher quality images.
The researchers have plans to push the neural network system further with ambitions to help transform shots in more drastic fashions, such as turning photos of a rainy backdrop into one bathed in sunny weather.
If this system finds its way into future smartphones it could boost the already impressive camera chops of flagship phones like Samsung’s Galaxy S8. At the moment, the best way to get AI-powered phone camera tech is to buy the Google Pixel 2 which uses the search giant’s machine learning prowess to produce some truly impressive smartphone photography.
Related: Best smartphones
What smartphone features would you like to see get a does of AI smarts? Let us know on Intertwine or Facebook.