3 ms·
I think a fully trained neural network is, in a sense, a "lossy" (possibly very lossy) description of the training set.
by Florin_Andrei 10y ago
I think a fully trained neural network is, in a sense, a "lossy" (possibly very lossy) description of the training set.
- Stanleyc23 10y agoOh what I was imagining was more like if I was watching a news broadcast and so all I care about was the news anchor person, and the main on screen grapic/text. If you Compress down other things like the table or the background or wrote an algorithm to selectively stream resolutions tied to the object streamed instead of the frame streamed. Is that a viable form of "compression" using existing computer vision tech?