Error in building the project

[metro-smiles]

Hi @chengzhag ,

I got the following error while building the project. Seems like there are some missing files.

Building pyrender...
Run: cd external/mesh_fusion/librender && /home/user/anaconda3/envs/pytorch11_py3_Im3D/bin/python setup.py build_ext --inplace
cc1plus: warning: command line option ‘-Wstrict-prototypes’ is valid for C/ObjC but not for C++
In file included from /home/user/anaconda3/envs/pytorch11_py3_Im3D/lib/python3.6/site-packages/numpy/core/include/numpy/ndarraytypes.h:1832:0,
                 from /home/user/anaconda3/envs/pytorch11_py3_Im3D/lib/python3.6/site-packages/numpy/core/include/numpy/ndarrayobject.h:12,
                 from /home/user/anaconda3/envs/pytorch11_py3_Im3D/lib/python3.6/site-packages/numpy/core/include/numpy/arrayobject.h:4,
                 from pyrender.cpp:531:
/home/user/anaconda3/envs/pytorch11_py3_Im3D/lib/python3.6/site-packages/numpy/core/include/numpy/npy_1_7_deprecated_api.h:17:2: warning: #warning "Using deprecated NumPy API, disable it with " "#define NPY_NO_DEPRECATED_API NPY_1_7_API_VERSION" [-Wcpp]
 #warning "Using deprecated NumPy API, disable it with " \
  ^
In file included from pyrender.cpp:534:0:
offscreen.h:4:21: fatal error: GL/glew.h: No such file or directory
compilation terminated.
error: command 'gcc' failed with exit status 1
Traceback (most recent call last):
  File "project.py", line 158, in <module>
    globals()[args.work](args)
  File "project.py", line 120, in build
    subprocess.check_output(cmd, shell=True)
  File "/home/user/anaconda3/envs/pytorch11_py3_Im3D/lib/python3.6/subprocess.py", line 356, in check_output
    **kwargs).stdout
  File "/home/user/anaconda3/envs/pytorch11_py3_Im3D/lib/python3.6/subprocess.py", line 438, in run
    output=stdout, stderr=stderr)
subprocess.CalledProcessError: Command 'cd external/mesh_fusion/librender && /home/user/anaconda3/envs/pytorch11_py3_Im3D/bin/python setup.py build_ext --inplace' returned non-zero exit status 1.

It seems there is no directory called GL. Hence, any reference to headers included in it would probably throw an error.

Would appreciate your help with this.

~Cheers!

[chengzhag]

Hi @metro-smiles,

It seems like that your OpenGL is not installed properly.

The installation of the libraries required should be done in

Implicit3DUnderstanding/external/build_gaps.sh

Line 37 in af2964f

sudo apt-get install mesa-common-dev libglu1-mesa-dev libosmesa6-dev libxi-dev libgl1-mesa-dev

when building the project, which should ask for user password and requires higher privilege.

Maybe check if you are running with a user account with the privilege to run apt with sudo?

[metro-smiles]

Hi @chengzhag,

Really appreciate your response. So it did install the graphics libraries after having asked for sudo privileges. So not sure what went wrong. I did not get any error message while the apt-get packages were installing.

[metro-smiles]

A quick follow up, was your system setup on Ubuntu 16 or 18?

[chengzhag]

Hi @metro-smiles,

Sorry for the inconvenience. My system is Ubuntu 18. I noticed that the error might be caused by the missing library freeglut3-dev and libglew-dev. Could you please try apt install freeglut3-dev libglew-dev as in https://github.com/chengzhag/Implicit3DUnderstanding/tree/af2964f074d941cd990cff81a9b5f75489586ed2/external/mesh_fusion ?

If the problem is solved, I will try to add this line to the project building.

Really appreciate your feedback!

[metro-smiles]

Hi @chengzhag,

Really appreciate your response.

So I have been trying to run on Ubuntu 16, so wonder if that's causing some incompatibility. But anyways, after installing your suggested packages now I'm getting a different error (from CUDA):

Building pyfusion...
Run: cd external/mesh_fusion/libfusiongpu && mkdir -p build && cd build && cmake .. && make && cd .. && /home/user/anaconda3/envs/pytorch11_py3_Im3D/bin/python setup.py build_ext --inplace
CMake Warning (dev) in CMakeLists.txt:
  No project() command is present.  The top-level CMakeLists.txt file must
  contain a literal, direct call to the project() command.  Add a line of
  code such as

    project(ProjectName)

  near the top of the file, but after cmake_minimum_required().

  CMake is pretending there is a "project(Project)" command on the first
  line.
This warning is for project developers.  Use -Wno-dev to suppress it.

CMake Error at /home/user/anaconda3/envs/pytorch11_py3_Im3D/share/cmake-3.16/Modules/FindPackageHandleStandardArgs.cmake:146 (message):
  Could NOT find CUDA (missing: CUDA_INCLUDE_DIRS) (found suitable version
  "9.0", minimum required is "6.5")
Call Stack (most recent call first):
  /home/user/anaconda3/envs/pytorch11_py3_Im3D/share/cmake-3.16/Modules/FindPackageHandleStandardArgs.cmake:393 (_FPHSA_FAILURE_MESSAGE)
  /home/user/anaconda3/envs/pytorch11_py3_Im3D/share/cmake-3.16/Modules/FindCUDA.cmake:1104 (find_package_handle_standard_args)
  CMakeLists.txt:34 (find_package)


Traceback (most recent call last):
  File "project.py", line 158, in <module>
    globals()[args.work](args)
  File "project.py", line 112, in build
    subprocess.check_output(cmd, shell=True)
  File "/home/user/anaconda3/envs/pytorch11_py3_Im3D/lib/python3.6/subprocess.py", line 356, in check_output
    **kwargs).stdout
  File "/home/user/anaconda3/envs/pytorch11_py3_Im3D/lib/python3.6/subprocess.py", line 438, in run
    output=stdout, stderr=stderr)
subprocess.CalledProcessError: Command 'cd external/mesh_fusion/libfusiongpu && mkdir -p build && cd build && cmake .. && make && cd .. && /home/user/anaconda3/envs/pytorch11_py3_Im3D/bin/python setup.py build_ext --inplace' returned non-zero exit status 1.

[chengzhag]

Hi @metro-smiles,

It looks like that libfusiongpu of mesh_fusion didn't find on your system. If data processing speed is not a problem, you can try to change 'libfusiongpu' in this line

Implicit3DUnderstanding/project.py

Line 103 in af2964f

f"cd {os.path.join(mesh_fusion_path, 'libfusiongpu')}",

into 'libfusioncpu'. Or try to install CUDA NVCC.

I'll add the requirement to README if it is the case.

[metro-smiles]

Hi @chengzhag ,

Really appreciate all your help/inputs! I was able to build and run the demo.

To summarize, following were the changes I ended having to make:

1. Install these libraries, separately:

$ apt install freeglut3-dev libglew-dev

2. Ensure nvcc is callable from command-line
3. If nvcc is not available, lifusioncpu can instead be used for build_gaps but nvcc will anyways be needed for ldif2mesh
4. Incase, gcc compiler is not set to C++ 2011 standard, set --std c++11 in the nvcc command in ldif2mesh/build.sh

[metro-smiles]

Hi @chengzhag ,

Apologies for re-opening the issue.

So while pre-processing the Pix3D dataset, it seems that the mesh_fusion/fusion.py code is not really using the GPU, eventhough I compiled the project with 'libfusiongpu' and set the use_gpu flag to True. Can you please suggest some possible explanation for this?

Btw, the demo component seems to work fine.

Appreciate your inputs.

[chengzhag]

Hi @metro-smiles,

From what I understand, the mesh_fusion/fusion.py does import libfusiongpu when use_gpu set to True. But further experiments indicate that GPU is truely not used (with the help of nvidia-smi) and there is no speed difference between libfusiongpu and libfusioncpu. It seems like that there is a problem with mesh_fusion lib, which I did not notice and cannot provide further help.

However, the preprocessing of Pix3D is accelerated with multi-thread, which can be done in less than hour. Thus the problem seems not very significant.

Thanks for your summary for the compiling problem. We updated two recent commits to solving it.

[metro-smiles]

Hi @chengzhag ,

Appreciate your inputs.

But it seems that the Pix3D pre-processing step is not as fast, at my end. I've tried it independently on two different systems but on both it's been running for over a day without making any progress. (It seems that call to fusion.py with mode=render is the most expensive and like I was saying before it's not using the GPU.)

Output of htop:

Is there any chance you pre-processed it differently / did some other intermediate step which is not mentioned in the README?

~Cheers!

[chengzhag]

Hi @metro-smiles,

I just recall that Pix3D pre-processing consists of two steps (i.e. image processing and mesh processing). Sorry that I wrongly think of the time consumption of image processing as the total pre-processing time.

Also my bad that I missed meshlab installation in the README, which you probably have found out or pre-installed.

As for the mesh processing step, I tried to run it with libfusiongpu, finding out that it actually has no problem using GPU:

I think It might be the case that the program opened too many processes for the machine to handle. Could you please try to run preprocess_pix3d4ldif with processes flag set to 0?
(Hint: You can comment out the lines between print('Processing imgs...') and print('Processing meshs...') to skip image processing.)