Releases: nihui/waifu2x-ncnn-vulkan
version 20200530
pre-compiled binary for windows / ubuntu / macos 64bit
build with ncnn 20200530
online spirv compilation, much smaller executable binary
link with static msvc runtime
webp image decoding
properly handle images with alpha channel
fix incorrect output with odd tilesize
reduce gpu memory usage
automatic tilesize settings according to the current gpu heap budget
github action ci added
version 20200414
pre-compiled binary for windows 64bit
vs2017
add fp16p executable for some old buggy driver
build with ncnn 20200414
version 20200224
pre-compiled binary for windows 64bit
vs2017
fix tta mode on odd-size image
fix cjk output
build with ncnn 20200224
version 20200223
pre-compiled binary for windows 64bit
vs2017
tta mode
fix image decoding encoding on win7
build with ncnn 20200223
version 20200222
pre-compiled binary for windows 64bit
vs2017
add upconv7-photo model
build with ncnn 20200222
version 20190712
pre-compiled binary for windows 64bit
vs2017
batch upscaling, directory to directory
new option for setting internal thread count of the three stages ( image decoding, waifu2x upscaling and image encoding )
build with ncnn 20190712, fix fp16p issue with radv driver on linux
version 20190630
pre-compiled binary for windows 64bit
vs2017
rename binary to waifu2x-ncnn-vulkan
command line option changes, new option for model dir and gpu device
add faster waifu2x model upconv_7_anime_style_art_rgb, aka UpRGB
version 20190611
pre-compiled binary for windows 64bit
vs2017
build with ncnn 20190611, faster conversion speed
version 20190421
pre-compiled binary for windows 64bit
vs2017
fix decode/encode error on filepath with non-ascii characters
version 20190414
pre-compiled binary for windows 64bit
vs2017
new command option for block size
rewrite tile processing logic to reduce gpu memory usage
faster host device copy via int8 storage shader