Skip to content

[Windows 11 Compatible] OpenPCDet Toolbox for LiDAR-based 3D Object Detection.

License

Notifications You must be signed in to change notification settings

Uzukidd/OpenPCDet-Win11-Compatible

 
 

Repository files navigation

[Windows 11 Compatible] OpenPCDet 0.6.0

This is a compatible version of OpenPCDet for WIndows 11.

OpenPCDet

Official README.md see here: README_aux.md

Problems known

  • Sharedarray removed, Waymo dataset with shared memory may not work.

Installation schme

  • Based on Global CUDA dev 11.4 installed
  • IoU3d_nms modified
  • Dependency Sharedarray removed

Environment tested

  • CUDA 11.4 + Py3.8 + VS2019 + Pytorch1.10

    • Windows 11

    • NVCC 11.4

    • Visual Studio 2019

    • Cudatoolkit 11.3

    • Anaconda 3

    • Python 3.8.16

    • Pytorch 1.10.1

    • spconv-cu113 2.3.6

    • NVIDIA GeForce RTX 3060 Laptop GPU

  • CUDA 11.8 + Py3.10 + VS2022 + Pytorch2.0

    • Windows 11

    • NVCC 11.8

    • Visual Studio 2022

    • Cudatoolkit 11.8

    • Anaconda 3

    • Python 3.10.11

    • Pytorch 2.0.0

    • spconv-cu118 2.3.6

    • NVIDIA GeForce RTX 3060 Laptop GPU

Changelog

[2023-05-1] Mayday united together

[2023-05-1] Dataset shared memory modified.

[2023-04-28] IoU3d_nms modified.

[2023-04-28] Readme and dev branch initialized.

Modification from official code

  • IoU3d_nms
  • Part of Dataset Shared Memory

How to use it

a. Install and configure CUDA 11.4 or CUDA 11.8 properly from https://anaconda.org/conda-forge/cudatoolkit-dev/files

b. Clone this repository.

git clone https://github.com/Uzukidd/OpenPCDet-Win11-Compatible.git

c. Create the environment as follows:

conda env create -f environment_cu114.yml (or environment_cu118.yml)

c. Install this pcdet library and its dependent libraries by running the following command:

python setup.py develop

ez check with PointPillar

a. Download easy_check from https://drive.google.com/file/d/13czqs5oSTb86QC0ZOMe6wEjgL91_k-VR/view?usp=sharing and extract it into ./tools/

b. Activate the environment

conda activate env_openpcdet_base

c. Enter tools

cd ./tools

d. Run the demo

python demo.py --cfg_file ./cfgs/kitti_models/pointpillar.yaml --ckpt ./easy_check/pointpillar_7728.pth --data_path  ./easy_check/000000.bin

About

[Windows 11 Compatible] OpenPCDet Toolbox for LiDAR-based 3D Object Detection.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 86.7%
  • Cuda 7.9%
  • C++ 4.8%
  • Other 0.6%