Researchers from ETH Zurich have developed a neural network to enhance images taken by standard smartphone cameras into "DSLR-quality" photographs.
The deep learning system, which is open for trial, only requires two distinct datasets – one from any generation of smartphone and other a DSLR level image, reports Engadget.
The content of the images may differ, but the system will process them in such a way that the quality of one image will be applied to the other, making it more vibrant with better colours and exposure.
The team that developed this system started out with a basic version, which was trained using photos taken of the same scene from a phone as well as a DSLR. It was limited to images taken from that particular smartphone.
However, the advanced version works even for dated smartphones limited by hardware capabilities. The modified images, as showcased by the team, may not always look that great. They might lose some detail or get a different tint, but that's still pretty impressive for improving dull images taken by low-end devices with weak cameras.
Moving ahead, the researchers plan on upgrading this tech by adding capabilities to tweak environmental conditions in a photograph. In other words, the system could have the capability to correct an image taken in rain to make it bright and sunny. While that's still a long way to go, incorporation of this system into upcoming phones could take image produced by low-end phones a bit closer to what is expected from devices with high-end sensors and lenses.