Quantcast

Neural Networks Provide A More Immersive Look at Early 1910s Japan

Japan Neural Network 1910s

A recent video is giving viewers a better look at Japan between the years 1913 and 1915, using old footage that has been “upscaled” via a neural network.

YouTuber Denis Shiryaev is known for using a neural network to upscale vintage film with color correction, improved resolution, and increased frames per second. The resulting footage is much more immersive than the original grainy film which is blurry, black and white, and only about twelve frames per second.

It is worth mentioning that the film’s enhancements aren’t authentic. While a neural network does its best to remaster the video, the added frames and increased resolution are filled in and may not represent the truth.

A neural network is a series of algorithms that engage in machine-based learning and can be trained to recognize patterns and generate new information. The most well known neural networks were those used in “Deep Fakes” which allowed users to superimpose someone else’s face in a piece of media.

In September 2019, DARPA (the Defense Advanced Research Projects Agency) reportedly began building technologies to detect fake images, videos, and news stories from “going viral”.

If the program is successful after four years of trials, it will expand to target all “malicious intent.” Tests include giving the program 500,000 stories- with 5,000 fakes among them.

However neural networks can also be used for things like remastering vintage film, creating music, add different actors to classic films, or making SpongeBob recite text-to-speech.

You can watch the remastered footage of early 20th century Japan below.

This is Nicchiban Culture. In this column, we regularly cover Japanese culture, geek culture, and things related to anime. Please leave feedback and let us know if there’s something you want us to cover!

, ,
Avatar

About

A basement-dwelling ogre, Brandon's a fan of indie games and slice of life anime. Has too many games and not enough time.