Drones are now widely used in filming or video making, but they are still unable to instruct drones, but it is now possible that a single word could enable drones to capture better scenes.
This work has been done by Rogerio Bonatti, a doctoral student at Carnegie Mellon University, and his colleagues, with the help of the Facebook AI program. They have created a drone model that captures the reactions of the viewers or videos with the desired emotions or feelings.
But how does it happen? Imagine that when you call a drone the words emotional, calm, work, enjoyable, and nerve-racking, it adjusts its speed, angles, flight path, and video accordingly. Compiles and captures scenes in the same way. In this way, directors can now direct the drone to film, and the drone acts on it.
For this, a sensor and processing power facility has been created in the drone, the entire system of which has been installed behind the camera. In this way, using artificial intelligence, he creates the best and most desired scenes. A summary of this research is presented at the 2021 International Robotics and Automation Conference.
"We've been thinking about controlling the drone camera with an emotion or a word, and now the dream has come true," Rogerio said. For this, scientists have seen thousands of videos that take the viewer into a special emotional state. All of these videos have been rated. Scientists then collected hundreds of selected videos in one place. In the next phase, thousands of people were shown 12 pairs of videos and told how they felt after watching the video.
Now a model has been created on this basis and the artificial intelligence program has been taught to make similar videos under the guidance. The drone shoots the same video when asked for such an interesting video.
In this way, videos of football matches and other occasions were made under the direction and they were shown to the spectators again and their reaction was known. In this way, drones have been able to make videos very close to human guidance.
0 Yorumlar