AI in Video Publishing platform
If You can build AI algorithms that detect this genre of pictures you will have a system that can filter out the videos that are interviews of two people talking.
Facial detection will extract the face of each of them and with these face pictures, you will be able to use google people to know their names or use another round way of getting information about them for this relationship graph that you are building and that we have already built.
You can go further using other kinds of videos of different genres
for example if it is the genre of movies taken from Hollywood movies or through National Geography. There are professional photographers shooting the movies and professional directors directing them. They all use conventions for framing their shot so you can exploit the shot framing like this
By understanding how shots are framed you will be able to optimize your facial detection algorithm and make it faster because you will know the regions of the pictures where the actors’ heads will more likely be.
If you want you can also develop a classifier that can tell the name of the shot framing of each picture of a movie. Some of the shot framing types used in filmmaking are there below.
With this information on hand, do you think that it is possible you build an AI system that can do Principal Component Analysis in order to detect the Directors or professional photographers that share the same filming style?
Once your algorithm works for detecting the kinds of videos of two persons discussing you can make it ameliorate it and make it able to detect videos of several persons talking like in this illustration.
There is an opportunity for a video platform that publishes only videos of people discussing not selfish videos or commercials ones and it will help to have these kinds of artificial intelligence system.
Go build it because I already did.