4 ms·
I don't know whether you noticed it, but the paper reveals it is co-authored by people from Adobe. There very well might be intentions towards that later on.
by Jhsto 10y ago
I don't know whether you noticed it, but the paper reveals it is co-authored by people from Adobe. There very well might be intentions towards that later on.
- ttoinou 10y agoThat might take years
- coldtea 10y agoAdobe has incorporated research stuff they've shown to the public as mere papers/demos into PS in just 1-2 versions several times.
- dagw 10y agoAdobe moves pretty fast when they want to. Take content aware fill for example. It went from a presentation at ACM SIGGRAPH to Adobe hiring some of the people who wrote the paper to a working feature in Photoshop in about a year.
- ttoinou 10y ago@dagwTrue @coldtea True However the technology is not ready yet. Takes too long to render* and needs too much VRAM (=> limited resolution, varying on the current GPU). But a consumer-grade plugin with an online renderer might be feasible for now. * Fast solutions include a pre-training step that can take days. Would be weird to ask the user to do this
- wyldfire 10y ago> Would be weird to ask the user to do this Alternatives to training include downloading the trained model. It could be a big download but an overnight download is not that weird to ask users to do.
- hoschicz 10y ago> Alternatives to training include downloading the trained model. It could be a big download but an overnight download is not that weird to ask users to do. Or keep the pretrained network on superpowerful cloud machines.