VideoHelp Forum

Our website is made possible by displaying online advertisements to our visitors.
Please consider supporting us by disabling your ad blocker or buy a VSO converter software :)
+ Reply to Thread
Page 1 of 2
1 2 LastLast
Results 1 to 30 of 45
Thread
  1. Very recently, there is a wave of interest among the Japanese community in developing a new algorithm for upscaling: waifu2x

    The basic idea of this algorithm was developed in the Chinese University of Hong Kong, but only recently a Japanese implemented it in LUA and uploaded it to GitHub:

    https://github.com/nagadomi/waifu2x
    (... and it starts getting viral)

    This original version use CUDA on Linux, and this seems to be problematic for many users...

    Then shortly afterwards, people starts porting it to Windows, and with GUI. These have no CUDA acceleration however.

    First Windows port (only upscaling, no training part):
    https://github.com/WL-Amigo/waifu2x-converter-cpp

    GUI All-in-one package:
    http://inatsuka.com/extra/koroshell/
    (Click on the floppy icon to download)

    AVX optimization:
    https://github.com/tanakamura/waifu2x-converter-cpp

    AVX2 optimization:
    https://github.com/logicmachine/waifu2x-opt

    To use the GUI, launch waifu2x_win_koroshell.exe, then drag-and-drop your image file on to the main window.

    Someone made an Avisynth port, but reportedly too slow to be of any use (0.02fps on Core i5):
    http://waifu2x-avisynth.sunnyone.org/

    I ported an OpenGL version to AviUtl:
    https://github.com/MaverickTse/aviutl-waifu2x-gl

    GUI CUDA Version (some what troublesome):
    https://github.com/lltcggie/waifu2x-caffe/releases/tag/1.0.0
    (download link at the bottom, ~130MB!)
    This GUI is Japanese-only. I have made another multilingual GUI that wraps it. For details see post 41:
    https://forum.videohelp.com/threads/372157-New-upscaling-algorithm-waifu2x?p=2423628&vi...=1#post2423628

    Integration of AVX/FMA/CUDA/OCL...:
    http://d.hatena.ne.jp/w_o/20150616#1434392833

    AviUtl Plugin that use the above DLL:
    https://github.com/YSRKEN/aviutl-waifu2x-w2xc
    The release package already contains the DLL.
    (DLL: http://d.hatena.ne.jp/w_o/20150619#1434643288)
    ↑At version 1.4.1, this should be the fastest AviUtl version if using with CUDA.
    @640x360 -> 1280x720 takes ~3 sec on GTX970OC(GigaByte)
    Note that earlier versions may have problem with CUDA acceleration.
    Click image for larger version

Name:	waifu-wxc141.PNG
Views:	12711
Size:	257.1 KB
ID:	32480


    VapourSynth Plugin(also use w2xc dll)
    https://github.com/HomeOfVapourSynthEvolution/VapourSynth-Waifu2x-w2xc/


    Here are some samples:
    Name:  zoneplate.png
Views: 18420
Size:  73.4 KBClick image for larger version

Name:	zoneplate(scale)(x2.000000).png
Views:	14774
Size:	698.8 KB
ID:	31985
    1. Fringe pattern. Left: Original 256x256 px. Right: Waifu2x double sized, no denoise. Quite accurate and no artifact IMO.

    Name:  Kotori 00.png
Views: 18290
Size:  190.9 KBClick image for larger version

Name:	Kotori 00_2x.png
Views:	14614
Size:	558.4 KB
ID:	31987
    2. Japanese-style animation image (a capture from the PC game, Da Capo). You can try the doubling the image using conventional bicubic, sinc or spline, you should notice the edges in the waifu2x images are a lot crisper.
    Last edited by MaverickTse; 15th Dec 2015 at 18:55. Reason: addition
    Stopping development until someone save me from poverty or get me out of Hong Kong...
    Twitter @MaverickTse
    Quote Quote  
  2. Member racer-x's Avatar
    Join Date
    Mar 2003
    Location
    3rd Rock from the Sun
    Search Comp PM
    It's very slow, but does a good job.

    I don't understand Japanese, but it's a simple matter of drag and dropping the image into the program window. It saves new image as a PNG in the same folder as original.
    Got my retirement plans all set. Looks like I only have to work another 5 years after I die........
    Quote Quote  
  3. Some explanation on the GUI:

    Click image for larger version

Name:	waifuGUI_01.png
Views:	13711
Size:	387.0 KB
ID:	31988
    Click image for larger version

Name:	waifuGUI_03.png
Views:	13611
Size:	382.7 KB
ID:	31989
    Click image for larger version

Name:	waifuGUI_02.png
Views:	13524
Size:	383.4 KB
ID:	31990
    Stopping development until someone save me from poverty or get me out of Hong Kong...
    Twitter @MaverickTse
    Quote Quote  
  4. For the CUDA-enabled GUI:
    Click image for larger version

Name:	waifu-caffee-illu-text.png
Views:	836
Size:	33.3 KB
ID:	31992

    For my multilingual GUI wrapper:
    Click image for larger version

Name:	waifu2x_multilang_gui.png
Views:	6812
Size:	36.4 KB
ID:	34859

    Which is pretty self-explanatory
    Last edited by MaverickTse; 15th Dec 2015 at 19:01.
    Stopping development until someone save me from poverty or get me out of Hong Kong...
    Twitter @MaverickTse
    Quote Quote  
  5. To use the cuDNN acceleration, you need to sign-up the CUDA developer program on Nvidia's site and install the cuDNN package... quite troublesome.
    It is fine to left out cuDNN IMO.
    Stopping development until someone save me from poverty or get me out of Hong Kong...
    Twitter @MaverickTse
    Quote Quote  
  6. I can't wait for a proper english translation of this gui and at least avi compression
    Quote Quote  
  7. You can use the AviUtl plugin for the time being, if you don't mind a slow 0.1~0.2fps for upscaling from NTSC/PAL to double size.
    Currently only the CUDA implementation may reach 1fps, but building that binary is as hard as building FFmpeg (and far less documentation).
    There is a CLI version for the CUDA one, and of course in English.
    Stopping development until someone save me from poverty or get me out of Hong Kong...
    Twitter @MaverickTse
    Quote Quote  
  8. Member racer-x's Avatar
    Join Date
    Mar 2003
    Location
    3rd Rock from the Sun
    Search Comp PM
    I'm not sure this would be a good option for video (too slow), but it is a great option for images. It would make a good free alternative to Genuine Fractals, for example.
    Got my retirement plans all set. Looks like I only have to work another 5 years after I die........
    Quote Quote  
  9. Member Cornucopia's Avatar
    Join Date
    Oct 2001
    Location
    Deep in the Heart of Texas
    Search PM
    All those examples seem to be of limited palette/anime. Any real-world/natural images to compare?

    Scott
    Quote Quote  
  10. Member racer-x's Avatar
    Join Date
    Mar 2003
    Location
    3rd Rock from the Sun
    Search Comp PM
    Here is a before / after on a jpg image I downloaded somewhere. The top one is the 2x upscaled version.
    Image Attached Thumbnails Click image for larger version

Name:	beachgaltheojuly.jpg
Views:	12158
Size:	63.5 KB
ID:	32039  

    Click image for larger version

Name:	beachgaltheojuly(scale)(x2.000000).png
Views:	3649
Size:	1.39 MB
ID:	32040  

    Got my retirement plans all set. Looks like I only have to work another 5 years after I die........
    Quote Quote  
  11. Member Cornucopia's Avatar
    Join Date
    Oct 2001
    Location
    Deep in the Heart of Texas
    Search PM
    Well, that's a very pleasant UPsize effect!

    Slow as it is for video, there are a few projects where that could REALLY help/fix things. Good to know, thanks.

    Scott
    Quote Quote  
  12. Banned
    Join Date
    Oct 2014
    Location
    Northern California
    Search PM
    Neural networks for upscaling....

    Okay, how is the network trained and how is it decided what looks go and what not for each training sample?
    Quote Quote  
  13. Member Cornucopia's Avatar
    Join Date
    Oct 2001
    Location
    Deep in the Heart of Texas
    Search PM
    Since very few of us here can read Japanese, that might be a little difficult to answer. Perhaps MaverickTse has some info...

    Scott
    Quote Quote  
  14. Banned
    Join Date
    Oct 2014
    Location
    Northern California
    Search PM
    Originally Posted by Cornucopia View Post
    Since very few of us here can read Japanese, that might be a little difficult to answer. Perhaps MaverickTse has some info...

    Scott
    Perhaps the noise suppression is neural...

    Tried the waifu2x-cafe, works on jpg, crashes on png.

    Last edited by newpball; 3rd Jun 2015 at 19:50.
    Quote Quote  
  15. Member racer-x's Avatar
    Join Date
    Mar 2003
    Location
    3rd Rock from the Sun
    Search Comp PM
    Works here on PNGs. It gives some warning box that I can't understand, but click OK and it goes to work.

    This program uses a seven step process, so it can take a long time depending on image size. The results are very good I think.
    Got my retirement plans all set. Looks like I only have to work another 5 years after I die........
    Quote Quote  
  16. DECEASED
    Join Date
    Jun 2009
    Location
    Heaven
    Search Comp PM
    And more importantly:
    even though the GUI works well under Windows XP,
    the command-line .EXE does not
    Quote Quote  
  17. Sorry guys.
    My main PC got a mysterious short-circuit and the new rig will not be available until 13th June
    (DDR4 RAM out-of-stock)

    Limited support until then.

    For how the algorithm actually works(in English), please visit the first GitHub link in the 1st post...
    I cannot really answer that myself since I do not know what the heck is "convolution".
    All i can tell is that everything works in YUV, and the heavy algorithm is only applied to Y. The UV planes are simple Bicubic resize.
    Any magnification not 2X or 4X will involve extra shrinking step(area average or nearest neighbor)... so upscaling to 1.6X is actually slower than 2X

    beware that on some images, waifu2x will create an excessively "oil-painting" feel, so that is why it is claimed to be designed for Anime-style images.

    You can train the algorithm using your own image set, but you need a Nvidia rig on Linux, since only the Linux version has the trainig component.

    Hopefully developers around the world will join the quest in making it faster...
    Stopping development until someone save me from poverty or get me out of Hong Kong...
    Twitter @MaverickTse
    Quote Quote  
  18. Neural networks are finally feasible? About goddamn time.
    This looks very promising, for animation at least.
    Quote Quote  
  19. Member
    Join Date
    Jun 2015
    Location
    Portugal
    Search PM
    I upscaled the first episode of Death Note to 960p.. I cant provide any links here, but you can surely find it somewhere if you search for it.
    Quote Quote  
  20. Upload a 30 second clip of it. I'd like to see it. How did you even convert a whole episode? It took me half a minute just to upscale ONE 480p image. A whole episode would take what, 10 days?
    Quote Quote  
  21. Member
    Join Date
    Jun 2015
    Location
    Portugal
    Search PM
    Originally Posted by -Habanero- View Post
    Upload a 30 second clip of it. I'd like to see it. How did you even convert a whole episode? It took me half a minute just to upscale ONE 480p image. A whole episode would take what, 10 days?
    It took me 16h for a full 22 min episode.

    2 sec per frame.

    If anyone is interested in a download link, pm me, since its against the rules to post it here.
    Quote Quote  
  22. Originally Posted by foxmlg View Post

    If anyone is interested in a download link, pm me, since its against the rules to post it here.
    AFAIR 10 sec is allowed even under most strict Hollywood regulations - select few 10 sec parts.
    I also try this one and i must say it works sometimes nice, sometimes not and yes it is slow but i don't expect speed from non optimized code - works more than good as a proof of concept.

    https://www.youtube.com/watch?v=CJn_jC4FNDo
    Quote Quote  
  23. Originally Posted by foxmlg View Post
    Originally Posted by -Habanero- View Post
    Upload a 30 second clip of it. I'd like to see it. How did you even convert a whole episode? It took me half a minute just to upscale ONE 480p image. A whole episode would take what, 10 days?
    It took me 16h for a full 22 min episode.

    2 sec per frame.

    If anyone is interested in a download link, pm me, since its against the rules to post it here.
    2sec per frame !!!
    which version did you use? and on what machine?
    Stopping development until someone save me from poverty or get me out of Hong Kong...
    Twitter @MaverickTse
    Quote Quote  
  24. Member
    Join Date
    Jun 2015
    Location
    Portugal
    Search PM
    Originally Posted by MaverickTse View Post
    Originally Posted by foxmlg View Post
    Originally Posted by -Habanero- View Post
    Upload a 30 second clip of it. I'd like to see it. How did you even convert a whole episode? It took me half a minute just to upscale ONE 480p image. A whole episode would take what, 10 days?
    It took me 16h for a full 22 min episode.

    2 sec per frame.

    If anyone is interested in a download link, pm me, since its against the rules to post it here.
    2sec per frame !!!
    which version did you use? and on what machine?

    This one: https://github.com/nagadomi/waifu2x

    On Ubuntu 14.04 64Bit, with a GTX 770.
    Quote Quote  
  25. The quality of your death note episode is not too good. It doesn't look any better than a substandard upscaler designed for cartoons.

    I'm guessing you used a horribly inefficient version of waifu2x that allowed it to be that fast. Send me one original frame of your video and I'll see if the waifi i'm using will do any better.
    Quote Quote  
  26. Member
    Join Date
    Jun 2015
    Location
    Portugal
    Search PM
    Originally Posted by -Habanero- View Post
    The quality of your death note episode is not too good. It doesn't look any better than a substandard upscaler designed for cartoons.

    I'm guessing you used a horribly inefficient version of waifu2x that allowed it to be that fast. Send me one original frame of your video and I'll see if the waifi i'm using will do any better.
    Here is some original frames: http://imgur.com/a/R8xiX
    And here is some upscaled frames: http://imgur.com/a/KHCTa

    I used the one I said, it uses CUDA acceleration, thats why its fast. I also upscaled it 2x with 2x denoise. Maybe thats why it doesnt look as good as you want.
    Also used the already trained waifu2x from the creator. Since I dont have enough ram to train it my self.
    I heard that now they trained it better, it may yield better results now.

    Also, if we were to train it extensively.. It would result in much better results..
    Quote Quote  
  27. Those photos are 580p, show me the 960p ones.
    Quote Quote  
  28. Member
    Join Date
    Jun 2015
    Location
    Portugal
    Search PM
    Originally Posted by -Habanero- View Post
    Those photos are 580p, show me the 960p ones.
    You must be blind, cus the first link is 480p, and the second link is 960p..
    Quote Quote  
  29. The second link shows 1024x580 photos. https://i.imgur.com/kKIja2ah.png
    Use a better image site.
    Quote Quote  
  30. Member
    Join Date
    Jun 2015
    Location
    Portugal
    Search PM
    Originally Posted by -Habanero- View Post
    The second link shows 1024x580 photos. https://i.imgur.com/kKIja2ah.png
    Use a better image site.
    Get a better computer, since the problem is with you.
    Quote Quote  



Similar Threads