experiences
Artificial Intelligence in Live Video
A few days ago at an event in Bucharest, I have presented some new technology solutions that can be utilised in live streaming. As SLS also deals with the building and operation of such systems, for example Askato, the key parts of this presentation are summarised below.

The novelty of these solutions is essentially that Artificial Intelligence (AI) places items in the video content that were not there before. These are generated by an AI software – not by a human. For example, the face of a person appearing in a video is replaced by another person’s face.

Well the really new thing is that this technology can now be used in a live streaming system, i.e. the faces of the chat partners can be replaced even in a live video. Until now, this seemed extremely difficult because of the huge server capacity demand, but the rapid development of AI has resolved this problem as well.
‍

In the video below, you can follow the transformation of the live video, using artificial intelligence. The first frame is a 2D image of Barack Obama’s face; the second is a video of an actor with similar facial proportions to that of Obama. The third frame shows the movement that AI understands from this, the fourth is the 3D surface that is used by the AI based on the image. The fifth is the finished result, an Obama interview that has never really happened (thanks for the video for Hao Li and and his company, Pinscreen).

Obviously, most of us get concerned about this technology seeing it the first time because it is a big question how we can now determine who we see in a video or who is sitting on the other side of a live video chat. Given that this technology is currently in the process of development, there are no solutions to these concerns for the time being. And as I have said in my presentation, any such solution would require an unprecedented degree of self-regulation and cooperation between technology companies and lawmakers.

Closing this entry on a high note, I present three possible uses of the technology – now focusing on the wider use of AI, besides face replacement –, which will make our life easier:

  • Entertainment: We can place unique filters on our home videos, such as snow and mountains on our holiday videos, or make us look years younger.
  • Privacy: We want to consult our doctor in a live video chat about a symptom, but we do not want our face to be recognized, let alone recorded, so we can put on a face randomly generated by AI.
  • Security: We can put an AI filter on a movie our child watches, which pixelates or masks violent content or certain body parts.
My other posts
experiences

Experimenting With Amazon Go

Amazon opened its first concept store to the public on January 22, 2018. According to a promotional video published by the company, the store uses several technologies, including computer vision, deep learning algorithms, and sensor fusion to automate much of the purchase, checkout, and payment steps associated with a retail transaction.

check post