mmdetection3d
mmdetection3d copied to clipboard
open-mmlab
[Bug] Recent change in local_visualizer.py causes error on Windows
Prerequisite
- [X] I have searched Issues and Discussions but cannot get the expected help.
- [X] I have read the FAQ documentation but cannot get the expected help.
- [X] The bug has not been fixed in the latest version (dev-1.x) or latest version (dev-1.0).
Task
I'm using the official example scripts/configs for the officially supported tasks/models/datasets.
Branch
main branch https://github.com/open-mmlab/mmdetection3d
Environment
sys.platform: win32 Python: 3.9.16 | packaged by conda-forge | (main, Feb 1 2023, 21:28:38) [MSC v.1929 64 bit (AMD64)] CUDA available: True numpy_random_seed: 2147483648 GPU 0: NVIDIA GeForce RTX 3080 Ti CUDA_HOME: C:\ProgramData\Anaconda3\envs\mmdetection NVCC: Cuda compilation tools, release 11.6, V11.6.124 GCC: n/a PyTorch: 1.12.0 PyTorch compiling details: PyTorch built with:
- C++ Version: 199711
- MSVC 192829337
- Intel(R) Math Kernel Library Version 2020.0.2 Product Build 20200624 for Intel(R) 64 architecture applications
- Intel(R) MKL-DNN v2.6.0 (Git Hash 52b5f107dd9cf10910aaa19cb47f3abf9b349815)
- OpenMP 2019
- LAPACK is enabled (usually provided by MKL)
- CPU capability usage: AVX2
- CUDA Runtime 11.6
- NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_61,code=sm_61;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75;-gencode;arch=compute_80,code=sm_80;-gencode;arch=compute_86,code=sm_86;-gencode;arch=compute_37,code=compute_37
- CuDNN 8.3.2 (built against CUDA 11.5)
- Magma 2.5.4
- Build settings: BLAS_INFO=mkl, BUILD_TYPE=Release, CUDA_VERSION=11.6, CUDNN_VERSION=8.3.2, CXX_COMPILER=C:/cb/pytorch_1000000000000/work/tmp_bin/sccache-cl.exe, CXX_FLAGS=/DWIN32 /D_WINDOWS /GR /EHsc /w /bigobj -DUSE_PTHREADPOOL -openmp:experimental -IC:/cb/pytorch_1000000000000/work/mkl/include -DNDEBUG -DUSE_KINETO -DLIBKINETO_NOCUPTI -DUSE_FBGEMM -DUSE_XNNPACK -DSYMBOLICATE_MOBILE_DEBUG_HANDLE -DEDGE_PROFILER_USE_KINETO, LAPACK_INFO=mkl, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, TORCH_VERSION=1.12.0, USE_CUDA=ON, USE_CUDNN=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=OFF, USE_MPI=OFF, USE_NCCL=OFF, USE_NNPACK=OFF, USE_OPENMP=ON, USE_ROCM=OFF,
TorchVision: 0.13.0 OpenCV: 4.7.0 MMEngine: 0.8.4 MMDetection: 3.0.0 MMDetection3D: 1.3.0+b84ba5a spconv2.0: True
Reproduces the problem - code sample
Reproduces the problem - command or script
python mmdetection3d/tools/misc/browse_dataset.py .\configs_base_\datasets\kitti-3d-3class.py --task lidar_det --output-dir .\data\kitti\ --show-interval 30
Reproduces the problem - error message
Traceback (most recent call last):
File "E:\mmdetection3d\tools\misc\browse_dataset.py", line 152, in
Additional information
Hello,
a recent change in the file local_visualizer.py requires that the environment variable DISPLAY exists in order to initialize the Open3d visualizer backend. Since Windows desktops do usually not set a DISPLAY variable, calls to the Det3DLocalVisualizer e.g. with the browse_dataset script, will terminate with the fatal exception shown above.
See: https://github.com/open-mmlab/mmdetection3d/blame/5c0613be29bd2e51771ec5e046d89ba3089887c7/mmdet3d/visualization/local_visualizer.py#L180
The browse_dataset will work normally again after either removing the if-condition or setting up a DISPLAY variable with any value. However, I think it would be good to rework the code section for system compatibility.
I have the same bug in Window platform. Do you know how to fix it? Thank you,
Now I got the same bug on Linux:
Traceback (most recent call last):
File "projects/BEVFusion/demo/multi_modality_demo.py", line 78, in
I've got the same bug, does anyone know how to fix this? Thanks!
Did you try out one of the workarounds that I suggested above?
For now, you could either try setting up an environment variable named DISPLAY or remove the if-clause in the code chunk that I linked. In my case, the error occurred because the following lines 181 and 182 were not executed if this variable is not present.
Be aware that setting up a DISPLAY environment variable might have undesired side effects with other programs that try to read it, so it's really just a workaround.
I've got the same bug, does anyone know how to fix this? Thanks!
Did you try out one of the workarounds that I suggested above?
For now, you could either try setting up an environment variable named DISPLAY or remove the if-clause in the code chunk that I linked. In my case, the error occurred because the following lines 181 and 182 were not executed if this variable is not present.
Be aware that setting up a DISPLAY environment variable might have undesired side effects with other programs that try to read it, so it's really just a workaround.
As you said, remove the if-clause is working! Appreciate for your help, Thank hou!
File "demo/pcd_demo.py", line 90, in
Thank you @MatthiasHinz for mentioning the issue and suggesting a workaround.
I followed your suggestion by commenting out line 180. However, I was having this error:
TypeError: paint_uniform_color(): incompatible function arguments. The following argument types are supported:
1. (self: open3d.cuda.pybind.geometry.LineSet, color: numpy.ndarray[numpy.float64[3, 1]]) -> open3d.cuda.pybind.geometry.LineSet
Invoked with: LineSet with 12 lines., 0.0
Additionally, I had to make some changes to the local_visualizer.py file following this PR #2729 to get rid of that error. Now I can visualize the point cloud with predicted 3D bounding boxes. This is my testing environment:
OS: Windows 11 GPU: NVIDIA GeForce RTX 4090 CUDA: 11.8 Python: 3.8 Pytorch: 2.1.2 TorchVision: 0.16.2 OpenCV: 4.9.0.80 Open3D: 0.18.0 MMEngine: 0.10.2 MMDetection: 3.3.0 MMDetection3D: 1.4.0 (Built from source)
Thank you @MatthiasHinz for mentioning the issue and suggesting a workaround.
I followed your suggestion by commenting out line 180. However, I was having this error:
TypeError: paint_uniform_color(): incompatible function arguments. The following argument types are supported: 1. (self: open3d.cuda.pybind.geometry.LineSet, color: numpy.ndarray[numpy.float64[3, 1]]) -> open3d.cuda.pybind.geometry.LineSet Invoked with: LineSet with 12 lines., 0.0Additionally, I had to make some changes to the file following this PR #2729 to get rid of that error. Now I can visualize the point cloud with predicted 3D bounding boxes. This is my testing environment:
local_visualizer.pyOS: Windows 11 GPU: NVIDIA GeForce RTX 4090 CUDA: 11.8 Python: 3.8 Pytorch: 2.1.2 TorchVision: 0.16.2 OpenCV: 4.9.0.80 Open3D: 0.18.0 MMEngine: 0.10.2 MMDetection: 3.3.0 MMDetection3D: 1.4.0 (Built from source)
Can I See your local_visualizer.py？ I still can't run it
Thanks for your reminder, we'll re-test on windows and fix this!
this bug still exists in mmdet3D v1.4.0 which is released recently
