The Beautiful Video Made When a Computer Predicts 100,000 Frames



Machine learning is an incredibly powerful thing. Damien Henry, a technical program manager at Google, took advantage of this, feeding a machine learning algorithm a single image and asking it to generate an hour-long video of approximately 100,000 frames by predicting the next frame based on the previous one. The result is gorgeous to watch.

Perhaps it’s my attachment to the piece of music (“Music for 18 Musicians,” a seminal minimalism piece by Steve Reich), but I found watching this to be strangely emotional. I think the beauty comes in that the fleeting tableaus are reminiscent enough of some pastoral to evoke memories of childhood — the imperfect rendering conjures the act of dreaming such things itself — yet at the same time, they are generic enough to be just as much your memory as they are mine. It’s also a powerful example of the future; entire scenes replete with detail — clouds, a beach, greenery, and more — appear all from the simple generating image shown below:

It raises interesting questions as well. Will this be a tool to augment photo and video work in the future? Will we able to fill in gaps in scenery with such technology? It’s exciting to watch this develop.

[via Gizmodo]



Original Source Link


Leave a Reply