>With “RAISR: Rapid and Accurate Image Super-Resolution”, we introduce a technique that incorporates machine learning in order to produce high-quality versions of low-resolution images. RAISR produces results that are comparable to or better than the currently available super-resolution methods, and does so roughly 10 to 100 times faster, allowing it to be run on a typical mobile device in real-time. Furthermore, our technique is able to avoid recreating the aliasing artifacts that may exist in the lower resolution image.
@
Thank you based Google!
>>57547136
Oh, i forgot the link:
https://research.googleblog.com/2016/11/enhance-raisr-sharp-images-with-machine.html
>>57547136
So this is basically just waifu2x trained on a different set of images.
How long until there's an mpv filter?
>>57547155
Looks like it, hah.
However,
>allowing it to be run on a typical mobile device in real-time
waifu2x is fairly slow, even with a ton of CUDA cores it takes around 1-5 seconds per frame.
Not sure what Google's definition is of "real-time" here though.
Since this is the only machine learning thread on /g/ right now, miles-deep just got updated. Scores and coverage should work now. All it needs is a toys and masturbation label and this shit is gold.
https://github.com/ryanjay0/miles-deep
>>57547927
The problem with strong NN upscalers like waifu is that the models are big and slow to evaluate as they work on the picture as a whole. This is the trick google uses:
The core idea behind RAISR is to enhance the quality of a very cheap (e.g. bilinear) interpolation method by applying a set of pre-learned filters on the image patches, chosen by an efficient hashing mechanism.
I think that NNEDI/NG1 actually does something slightly similiar but weakly - being video deinterlacers/upscalers, it can "train" layers from local temporal domain of past frames (ie model is built on the fly).
>>57550029
>Be pretty smart amd understand machine learning
>Better use my skillset to categorise scenes of porn so i can view porn with maximum efficiency