Restoring color for 100 year old videos with AI

Black-and-white video of Chinese living scenes recorded by a Canadian photographer since 1920 has been color-restored using AI and machine learning. This machine learning system was researched and developed by DGSpitzer , a Chinese developer.

In order to restore color for these hundred-year-old black-and-white footage, DGSpitzer has combined a number of AI tools and machine learning, including:

  1. DeOIdify : A machine learning system that adds color to black and white scenes.
  2. DAIN: Algorithm to add frames to videos, aiming to reproduce videos at 60 frames per second.
  3. ESGRAN : Another AI system, used to increase video resolution.
  4. DeOIdify : Another system that restores music to video.

Picture 1 of Restoring color for 100 year old videos with AI
Black and white video recording the scenes of the Chinese people were restored.

Adding frames and colors makes the centenarian's video feel more authentic and closer. However, the results are not really perfect because the images in the video are not really sharp, especially in the edges of images of people and objects are quite blurred.

DGSpitzer said the project was based on the guidance of Denis Shiryaev, who used AI tools to restore 4K 60 FPS to the classic silent film 'Arrival of a Train at La Ciotat Station - The train to La Station. Ciotat ' debuted in 1896.