Hello,

I am trying to compile the opencv_src with your modifications to use it with python but unfortunately I can't. I first tried to build opencv from source but got an error then I also try to use the build_opencv.sh file you provide to see if I could build the wasm version but also got an error.

Compiling imgalign/opencv_src:

I followed the basic flow for building opencv from source (creating empty build folder in opencv_src, getting into it, running cmake .. and running make -j 4 ). At the cmake .. step I got a loop dependency error

CMake Error at cmake/OpenCVModule.cmake:432 (message):
  FATAL: Unresolved dependencies or loop in dependency graph (2)

  Processed OPENCV_MODULE_opencv_calib3d_DEPS:
  opencv_calib3d;opencv_core;opencv_features2d;opencv_flann;opencv_imgproc

  Good modules: opencv_core;opencv_flann;opencv_imgproc

  Bad modules: opencv_calib3d;opencv_features2d
Call Stack (most recent call first):
  cmake/OpenCVModule.cmake:601 (__ocv_sort_modules_by_deps)
  cmake/OpenCVModule.cmake:368 (__ocv_resolve_dependencies)
  modules/CMakeLists.txt:7 (ocv_glob_modules)

After looking at imgalign/opencv_src/modules/calib3d/CMakeLists.txt and imgalign/opencv_src/modules/features2d/CMakeLists.txt it looks like they both depend on each other.

I looked at the official opencv github repository and saw that the features2d module does not depend from calib3d module so I removed the dependency from your code and the cmake command ran but the make command is failing

...
In file included from /imgalign/opencv_src/modules/features2d/src/img_align/Display.cpp:2:
/imgalign/opencv_src/modules/features2d/src/img_align/Display.h:45:19: note: shadowed declaration is here
   45 |   TConstImages&   images;
      |                   ^~~~~~
In file included from /usr/include/c++/9/vector:67,
                 from /imgalign/opencv_src/modules/core/include/opencv2/core/types.hpp:53,
                 from /imgalign/opencv_src/modules/core/include/opencv2/core.hpp:58,
                 from /imgalign/opencv_src/modules/features2d/include/opencv2/features2d.hpp:47,
                 from /imgalign/opencv_src/build/modules/features2d/precomp.hpp:46:
/usr/include/c++/9/bits/stl_vector.h: In instantiation of 'class std::vector<const cv::Mat>':
/imgalign/opencv_src/modules/features2d/src/img_align/Display.cpp:46:29:   required from here
/usr/include/c++/9/bits/stl_vector.h:398:66: error: static assertion failed: std::vector must have a non-const, non-volatile value_type
  398 |       static_assert(is_same<typename remove_cv<_Tp>::type, _Tp>::value,
      |                                                                  ^~~~~
/usr/include/c++/9/bits/stl_vector.h:401:64: error: static assertion failed: std::vector must have the same value_type as its allocator
  401 |       static_assert(is_same<typename _Alloc::value_type, _Tp>::value,
      |                                                                ^~~~~
...

2. Building with build_opencv.sh

To build the wasm version I install emscripten following the instructions here https://emscripten.org/docs/getting_started/downloads.html. After I had to define the environment variable export EMSCRIPTEN=imgalign/emsdk/upstream/emscripten and by running bash build_opencv.sh I get the following error:

...
-- Performing Test HAVE_CPU_BASELINE_FLAGS
-- Performing Test HAVE_CPU_BASELINE_FLAGS - Failed
CMake Error at cmake/OpenCVCompilerOptimizations.cmake:507 (message):
  Compiler doesn't support baseline optimization flags:
Call Stack (most recent call first):
  cmake/OpenCVCompilerOptions.cmake:270 (ocv_compiler_optimization_options)
  CMakeLists.txt:555 (include)

Can you help me understand why I can't build your opencv version please?

Here is my configuration :
Ubuntu 20.04
cmake version 3.16.3
gcc 9.4.0
emsdk 3.1.7
(I succeed to build the opecnv from source on branch 3.4 with this config)

建议在docker环境下安装,因为不会干扰已有的环境.

1. sudo apt-get install npm
2. npm install
3. npm run build &
4. npm run serve &

注意:
在docker容器内开启的时候,实例化时需要指定端口.

例如:
docker run -p 8082 --name test_http -it docker_images_name:v1.0 /bin/bash

参考链接:

1. https://www.jianshu.com/p/9bc970082020

@latsic, multi-stitcher fails with in some instances when given more 4-5 images, and the output stitches only 3 images. Just wondering, is there anything I need to look into? Thanks!

Great app! I have a question around how to set params.

1. I understand that Matcher tries to accommodate distortions, but only some distortions are realistic. If the second image is low res, or the target is small, the algorithm returns a very distorted aligned image, when a simple rotated and scaled image would be a better fit. Is there a param setting that would prevent matches from returning a distorted image? Maybe a limit on the transform matrix? Or a switch to return "0" for success when the transform is deemed unlikely (as per the verbose logging)?

I have no web dev experience, I would like to use the backend code. Can you please help?

Thanks!

Hello,

If I try to create a panorama of my room and of my friend's room, it seems that the algorithm doesn't recognise a lot of the photographs giving the error "No match found for images with index: [list]" .

Is there a specific way required to take the photos?

Thanks,

Currently using the multistitcher in browser. Works very well.

Question about the resolution though. Are my input images being downsampled, or is the final image being reduced to less than its full potential resolution? Any way to change the settings to maximize the size of the final render?

Hey,
Thank you for creating this awesome repo!
As I would like to compile with the current version of OpenCV I would like to understand what modifications where necessary - could you elaborate on that? Where there any tutorials that you used?

I also have the issue that, when using the build_opencv.sh within a docker image, the resulting script is different and does not contain the ImgStitch function. Did you compile using a docker image? Which version of emscripten did you use?
I used: docker run --rm -v `pwd`:`pwd` -it trzeci/emscripten:sdk-tag-1.38.31-64bit /bin/bash