Researchers at the University of Washington have succeeded in generating a video with a single photo, and without any other image as a reference, through machine learning; the intelligent learning capacity of a neural network.
This is an automated image processing process that does not require additional reference material, and allows generating new images that fit with the one originally provided until enough images are generated to obtain a video that shows movement where there was none in the original material.
The procedure has some similarities in terms of the result with respect to what MyHeritage recently achieved; giving movement to old black and white photos (although it can obviously also be applied to a recent photograph), making the ancestors can «come back to life» as animated GIFs.
The researchers have achieved a striking result with the movement of natural elements such as clouds, water or smoke, which come to life in a completely realistic way in the middle of a photograph in which the rest of the landscape appears static.
Machine Learning
Like any machine learning process, this one, carried out by the Paul G. Allen School of Computer Science and Engineering at the University of Washington, begins with the training of a neural network fed with hundreds of videos of rivers, waterfalls and even seas and oceans. So that it could learn by itself the basics of the spectrum of moving bodies of water.
At first, learning is based on predicting, starting from a video, what a subsequent frame will look like.
As in this case the frame is available, the neural network is able to check its success rate and correct errors to improve on the next attempt, so that it is gradually able to produce results that are closer and closer to reality.
The next step is to display a static image from which to work out what the next frame would be; based on the pattern of water mass movements deduced.
At the moment the result of this machine learning process is very realistic and only small details (such as the refraction of light at certain points) can give clues that it is not a real image. However, everything can be corrected by increasing the learning period and providing the neural network with even more videos of moving bodies of water.
Continue reading: Sharing tweets on Instagram is now a fact without taking screenshots
We invite you to follow us on our social networks:
Instagram: @Creatigrafca
Facebook: Creatigraf.c.a
Twitter: @Creatigrafca
Youtube: @Creatigrafca