+1 vote
Is there even such an idea? That is, using a neural network to recreate the entire scene from the video sequence, for example the scene where the two are talking in the room? Question 2, is it possible to decompose the movie into a code with this operation, and compress it as a result.
Theoretically it is definitely possible, but at the moment as far as I know all attempts to implement it have not been successful. I will not give any proofs, I just do not remember where I read about it. In my opinion, the most promising direction for video compression using ANN are convolutional ANNs augmented at the decompression stage with ANNs that can increase the detail (the notorious demo )

1 Answer

0 votes
It is possible to partially reconstruct the scene. https://en.wikipedia.org/wiki/3D_reconstruction