Release 20250915

update ncnn libjpeg-turbo zlib-ng

Release 20250802

update ncnn and pre/post precess

• insanely faster on amd rdna3+
• faster on intel / amd integrated graphics
• slightly faster on nvidia
• less VRAM on old graphics without fp16 ssbo capability

image decoder / encoder with libjpeg-turbo / libpng / zlib-ng

• faster image loading and saving with simd optimization
• support very large resolution

linux package build for manylinux_2_28

expected to be compatible with other distros using glibc 2.28 or later, including:

• Debian 10+
• Ubuntu 18.10+
• Fedora 29+
• CentOS/RHEL 8+

Release 20250504

fix cpu mode with tilesize smaller than image size (#187)

It was only tested on Linux platform. So, please check if it works on Windows. :)

Signed-off-by: ArchieMeng <archiemeng@protonmail.com>

Release 20220728

update ncnn

Release 20220419

fix macos model directory
fix palette png loading in windows
update ncnn library for faster inference

Release 20210521

build smaller binary size

Release 20210210

update scale implementation

version 20210102

support Apple M1
support running on cpu, which can be parallelized with gpu, via -g -1

version 20200818

fix grayscale image decoding
save to jpg format
try looking for model files from executable directory
add cmake option to build with system libncnn and libwebp
update ncnn for faster gpu loading
multi-gpu support

version 20200606

webp encoding
fix color on windows