feigechuanshu / gfpgan-ncnn Goto Github PK

Hi. Thank You for your work.
Could you help me convert new GFPGANv1.3.pth model to onnx format?
Because I get errors in convert process convering GFPGANv1.3.pth to ONNX

Comiler Mingw-w64
here log with errors

[100%] Linking CXX executable Z:\AI_SDK\CPP_GFPGAN\GFPGAN-ncnn-main\ncnn_gfpgan.
exe
"C:\Program Files\CMake\bin\cmake.exe" -E cmake_link_script CMakeFiles\ncnn_gfpg
an.dir\link.txt --verbose=1
"C:\Program Files\CMake\bin\cmake.exe" -E rm -f CMakeFiles\ncnn_gfpgan.dir/objec
ts.a
C:\mingw64\bin\ar.exe qc CMakeFiles\ncnn_gfpgan.dir/objects.a @CMakeFiles\ncnn_g
fpgan.dir\objects1.rsp
C:\mingw64\bin\x86_64-w64-mingw32-g++.exe -std=c++11 -pie -fPIE -fPIC -Wall -O3
-fopenmp -O3 -DNDEBUG -Wl,--whole-archive CMakeFiles\ncnn_gfpgan.dir/objects.a
-Wl,--no-whole-archive -o Z:\AI_SDK\CPP_GFPGAN\GFPGAN-ncnn-main\ncnn_gfpgan.exe
-Wl,--out-implib,Z:\AI_SDK\CPP_GFPGAN\GFPGAN-ncnn-main\libncnn_gfpgan.dll.a -Wl,
--major-image-version,0,--minor-image-version,0 @CMakeFiles\ncnn_gfpgan.dir\link
libs.rsp
C:/mingw64/bin/../lib/gcc/x86_64-w64-mingw32/12.1.0/../../../../x86_64-w64-mingw
32/bin/ld.exe: CMakeFiles\ncnn_gfpgan.dir/objects.a(gfpgan.cpp.obj):gfpgan.cpp:(
.text+0x13e): undefined reference to ncnn::Net::Net()' C:/mingw64/bin/../lib/gcc/x86_64-w64-mingw32/12.1.0/../../../../x86_64-w64-mingw 32/bin/ld.exe: CMakeFiles\ncnn_gfpgan.dir/objects.a(gfpgan.cpp.obj):gfpgan.cpp:( .text+0xcc0): undefined reference to ncnn::Mat::reshape(int, int, int, ncnn::Allocator*) const'

Hello, I can't download the models you made from baidu website because I don't live in China, please put them somewhere so people can download easier as well.

could this work on android , inference on android device?

How to compile for Windows?

你好，感谢你的工作
我是一名ncnn初学者，我在mac上做了简单修改，成功运行了你的代码
现在我想学习下如何ncnn模型文件是如何转成功的呢？

我目前的步骤为：
1、按如下修改GFPGan的stylegan2_clean_arch.py 中的代码

        # noise injection
        if noise is None:
            b, _, h, w = out.shape
            # noise = out.new_empty(b, 1, h, w).normal_()
            # noise = torch.empty(b, 1, h, w).cuda()
            # nn.init.normal_(noise)
            noise = torch.randn(b, 1, h, w).cuda()

2、将GFPGANCleanv1-NoCE-C2.pth 转为了onnx
3、然后应用了python -m onnxsim model.onnx model-sim.onnx
4、./onnx2ncnn model-sim.onnx encoder.param encoder.bin
报错

Gather not supported yet!
 \# axis=1
zsh: segmentation fault  ./onnx2ncnn model-sim.onnx encoder.param encoder.bin

OS: Windows 10 21H1
Build tool: Visual Studio 2019

Commands:

mkdir build
cd build
cmake ..
cmake --build .

Result:

LINK : fatal error LNK1104: 无法打开文件“ncnn.lib” [D:\GFPGAN-ncnn\build\ncnn_gfpgan.vcxproj]