Neural network taught to remove people from real-time video

Recent AI experiments have proven that neural networks have enormous potential. Smart algorithms learn to replace faces in films and games. On this, the capabilities of neural networks are not limited. Now everyone can remove a person from real-time video.

Browser code from enthusiast Jason Mayes was discovered on GitHub. He created a “smart” algorithm based on the TensorFlow.js neural network, which he designed independently. The program has a unique ability: to remove a human figure from a video in real time.

The algorithm works on the basis of background recognition with the completion of the missing parts based on the information already received. This content requires JavaScript to run. The result is not always perfect. Often, the work of AI is visible to the naked eye. The developer writes that this is just an experiment. He does not say whether he will refine his algorithm. In any case, all sources are available for download and can be picked up by other enthusiasts.

Now anyone can check the operability of the neural network via the link, giving the site access to the camera of their device.

Source: github.com