A breakthrough in photography: Google and software developers at MIT created a program that edits photos before the shutter button is pressed. The software was developed using feedback from five photographers on 5,000 photos, and it continues to determine further adjustments through machine learning and artificial intelligence.
The software lets photographers see the final result, including the necessary edits, before taking the photo. That is a welcome development for amateur photographers and a benefit for professionals, who can work more easily, quickly, and effectively.
Using a Neural Network
The MIT and Google software developers used a neural network. Programmers had previously developed a similar neural-network program, but its computing requirements made it impractical. The earlier version needed more than 12 gigabytes of memory. The latest software requires only 100 megabytes, making it easier to use on a wider range of devices.
A Smartphone Chip Already Enables AI
People who enjoy taking photos with a smartphone can already use the ARM smartphone chip, which uses artificial intelligence to make adjustments and changes. The chip is based on machine learning, enabling a smartphone to perform complex tasks independently.
This closely reflects the growing capabilities of neural networks and artificial intelligence, which can substantially improve software quality without increasing computing requirements.
These expanded capabilities fit the broader development of data distribution and sharing. Virtual tours of homes for sale were already possible when this article was published, and artificial intelligence was increasingly allowing computers and devices to perform tasks independently. Similar developments could be seen in smart mobility, applications that communicate with users, and the live casino sector, where livestreaming technology increasingly relied on image quality and speed. At the time of publication, TensorFlow was expected to help accelerate comparable advances on smartphones.
TensorFlow and Artificial Intelligence in Photography
TensorFlow is also a neural-network system. This Google program can make image files considerably smaller without sacrificing quality. At the time of publication, it was still in its infancy and far from perfect. Combining TensorFlow's capabilities with a neural network's artificial intelligence—to preview edited photos before they are taken—opened up many interesting possibilities for near-term development.
The introduction of neural networks and artificial intelligence enables new features in photography and filmmaking, allowing computers and people to work together toward the best possible result.
