dockerfile

[hoenigpeter]

Hi,

great work implementing this pose estimation architecture and sharing it with us.
However I have troubles getting it running due to dependency issues and so on
Is there a dockerfile available to get it running quickly?

Help is appreciated

[shanice-l]

Sry I'm not very familiar with docker. Can you make a docker file and share it with us?

[francescosarno]

If you can provide a docker file, it would be great

[hoenigpeter]

@francescosarno @shanice-l you can find my dockerfile as well as docker-compose file in the new pull request

[francescosarno]

Thank you for your help! I used your dockerfile and it works perfectly. But when I try to run the training of pose estimation I got the following error:

from . import CppEGLRenderer ImportError: cannot import name 'CppEGLRenderer' from 'lib.egl_renderer' (/gdrnpp_bop2022/core/gdrn_modeling/../../lib/egl_renderer/__init__.py)

Do you know what is this caused by?

[hoenigpeter]

@francescosarno in the docker container, you need to compile the CppEglRenderer with:
bash lib/egl_renderer/compile_cpp_egl_renderer.sh

[NoelLD]

Error: Bindless Textures not supported

Hello, thank you very much for contributing a working docker container. Testing the pose estimation worked fine for me, tested for lmo and itodd datasets. But I'm running into an error when I try training the pose etimation (tested for lmo and itodd dataset).

example: sudo ./core/gdrn_modeling/train_gdrn.sh configs/gdrn/lmo_pbr/convnext_a6_AugCosyAAEGray_BG05_mlL1_DMask_amodalClipBox_classAware_lmo.py 0

Error:RuntimeError: Bindless Textures not supported

The error is raised by Python Assert in lib/egl_renderer/glutils/egl_offscreen_context.py (line: 225-226): if not glInitBindlessTextureNV():

The function OpenGL.GL.NV.bindless_texture.glInitBindlessTextureNV() - (Returns a boolean indicating whether this extension is available: GL_NV_bindless_texture)

I checked the OpenGL extensions with the OpenGL Extensions Viewer on my host system, but the GL_NV_bindless_texture is listed.
I also installed the components by using the bash scripts/compile_all.sh. Some components threw an error but building the cpp egl renderer worked.
NVidia driver and toolkit is installed and accessible from inside the docker container.
Vcxsrv is used as a xserver, installed and tested.

Does anyone have the same problem or have an idea how to solve it?

additional Details:

• Host OS: Windows 11
  • using WSL 2
• GPU: NVIDIA GeForce RTX 2060
  • (Extension GL_NV_bindless_texture existing, checked)
  • Version: 4.6.0 NVIDIA 536.40

I also tried the run the Container on a different computer with Ubuntu as host OS, but the container didn't seem to start (maybe because it's an old one with too little ram.

Full Error output:

20230823_152019|DBG|OpenGL.platform.ctypesloader@65: Loaded libEGL.so => libEGL.so <CDLL 'libEGL.so', handle 7956b70 at 0x7fbae259d250>
20230823_152019|ERR|__main__@233: An error has been caught in function '<module>', process 'MainProcess' (16), thread 'MainThread' (140442637346624):
Traceback (most recent call last):

 File "./core/gdrn_modeling/main_gdrn.py", line 233, in <module>
    main(args)
    │    └ Namespace(config_file='configs/gdrn/itodd_pbr/convnext_a6_AugCosyAAEGray_BG05_mlL1_DMask_amodalClipBox_classAware_itodd.py', ...
    └ <function main at 0x7fba8b9ab3a0>

  File "./core/gdrn_modeling/main_gdrn.py", line 199, in main
    Lite(
    └ <class '__main__.Lite'>

  File "/usr/local/lib/python3.8/dist-packages/pytorch_lightning/lite/lite.py", line 408, in _run_impl
    return run_method(*args, **kwargs)
           │           │       └ {}
           │           └ (Namespace(config_file='configs/gdrn/itodd_pbr/convnext_a6_AugCosyAAEGray_BG05_mlL1_DMask_amodalClipBox_classAware_itodd.py',...
           └ functools.partial(<bound method LightningLite._run_with_strategy_setup of <__main__.Lite object at 0x7fbb595f7c10>>, <bound m...

  File "/usr/local/lib/python3.8/dist-packages/pytorch_lightning/lite/lite.py", line 413, in _run_with_strategy_setup
    return run_method(*args, **kwargs)
           │           │       └ {}
           │           └ (Namespace(config_file='configs/gdrn/itodd_pbr/convnext_a6_AugCosyAAEGray_BG05_mlL1_DMask_amodalClipBox_classAware_itodd.py',...
           └ <bound method Lite.run of <__main__.Lite object at 0x7fbb595f7c10>>

  File "./core/gdrn_modeling/main_gdrn.py", line 155, in run
    renderer = get_renderer(cfg, data_ref, obj_names=train_obj_names, gpu_id=render_gpu_id)
               │            │    │                   │                       └ 0
               │            │    │                   └ ['1', '2', '3', '4', '5', '6', '7', '8', '9', '10', '11', '12', '13', '14', '15', '16', '17', '18', '19', '20', '21', '22', '...
               │            │    └ <module 'ref.itodd' from '/gdrnpp_bop2022/core/gdrn_modeling/../../ref/itodd.py'>
               │            └ Config (path: configs/gdrn/itodd_pbr/convnext_a6_AugCosyAAEGray_BG05_mlL1_DMask_amodalClipBox_classAware_itodd.py): {'OUTPUT_...
               └ <function get_renderer at 0x7fba8bdc2dc0>

  File "/gdrnpp_bop2022/core/gdrn_modeling/../../core/gdrn_modeling/engine/engine_utils.py", line 281, in get_renderer
    ren = EGLRenderer(
          └ <class 'lib.egl_renderer.egl_renderer_v3.EGLRenderer'>

  File "/gdrnpp_bop2022/core/gdrn_modeling/../../lib/egl_renderer/egl_renderer_v3.py", line 81, in __init__
    self._context = OffscreenContext(gpu_id=cuda_device_idx)
    │               │                       └ 0
    │               └ <class 'lib.egl_renderer.glutils.egl_offscreen_context.OffscreenContext'>
    └ <lib.egl_renderer.egl_renderer_v3.EGLRenderer object at 0x7fba8b93a8b0>

  File "/gdrnpp_bop2022/core/gdrn_modeling/../../lib/egl_renderer/glutils/egl_offscreen_context.py", line 156, in __init__
    self.init_context()
    │    └ <function OffscreenContext.init_context at 0x7fbae202dca0>
    └ <lib.egl_renderer.glutils.egl_offscreen_context.OffscreenContext object at 0x7fba8b8e5250>

  File "/gdrnpp_bop2022/core/gdrn_modeling/../../lib/egl_renderer/glutils/egl_offscreen_context.py", line 228, in init_context
    raise RuntimeError("Bindless Textures not supported")

RuntimeError: Bindless Textures not supported

[CJSUN997]

I meet the same error in WSL2 docker environment. And I think this is because bindlessTexuture maybe only supported by high version OpenGL or not supported by mesa deriver in wsl.
Have you tried to comment glInitBindlessTextureNV() in egl_offscreen_context.py?