Lately, we can read the news about a new AI achievement almost every day.
Just a few days ago I was writing about an AI model that can fix and color old black and white movies (read), and now I’m writing about something similar.
Namely, AI developer and Youtube Denis Shiryaev used "several neural networks" to scale and improve the famous train scene from 1895 to 4K resolution at 60fps!
It is a short French documentary called "L'arrivée d'un train en gare de La Ciotat" (Train arrives at the station) which shows the train docking at Ciotat station. The film is known for allegedly causing panic among moviegoers because they feared the train would get off the screen. Although these rumors were dismissed as inaccurate, you can imagine how people reacted at the time when they first saw the moving image on the canvas.
According to his Reddit post, Shiryaev "employed several neural networks" to scale the film to 4K at 60fps. Below you can see the original image, and below it is a scaled 4K image.
As you can see, AI has removed all those irregularities in the recording that are otherwise very visible in old movies. Of course, if you take a closer look, you can see new irregularities generated by the AI due to the lack of information in those parts of the image.
The sound was added to the recording, which certainly had a positive effect on the overall impression of this remastered work.
However, it is not yet known which neural networks Shiryaev used, but it is probably GAN (generative adversarial networks). These are the same networks used to create Deepfakes and FaceGAN.
Manipulating videos with the help of artificial intelligence can be dangerous for all of us because the results are very real. But, as we can see, AI is also used for good purposes, and that's the way it should be.
Comments
Post a Comment