4 ms·
In the example image the compression seems more noticeable on some parts of the image than others, though this is probably due in part to my imagination. It giv
by bbctol 10y ago
In the example image the compression seems more noticeable on some parts of the image than others, though this is probably due in part to my imagination. It gives me a wacky (far-future) idea, though; what about combining a NN-based compression system with data on eye tracking and attention, to vary the level of detail in an image depending on salience?
- IgorPartola 10y agoHah. Porn would be a prime candidate for compression like that.
- varikin 10y agoI believe John Carmack has discussed this idea in regard to VR. I haven't actually tried current VR or tracked the development that closely to say whether this is currently feasible or not, but VR with eye tracking is perfect place for optimization. Why have high res textures and such pushed out to what you are not looking at.