深度学习激光3D目标检测OpenPCDet安装笔记

安装环境

1. python=3.7
2. pytorch=1.3.1
3. torchvision=0.4.2
4. cudatoolkit=10.1

具体执行步骤：

conda create -n pytorch-1.3.1 python=3.7（创建环境）
conda activate pytorch-1.3.1（进入环境）
conda install pytorch==1.3.1 torchvision==0.4.2 cudatoolkit=10.1 -c pytorch（安装环境）

执行第三步时，因为pytorch-1.3.1-py3.7_cuda10.1.243_cudnn7.6.3_0.tar.bz2在线安装比较慢，我下载后离线安装：

conda install --offline pytorch-1.3.1-py3.7_cuda10.1.243_cudnn7.6.3_0.tar.bz2

然后重新执行命令，即可完成安装。

conda install pytorch==1.3.1 torchvision==0.4.2 cudatoolkit=10.1 -c pytorch

OpenPCDet demo运行

1. 进入到/home/higer/apps/OpenPCDet/tools目录

   cd tools/
   

2. 执行domo命令

   python demo.py --cfg_file cfgs/kitti_models/pv_rcnn.yaml --ckpt pv_rcnn_8369.pth --data_path ../data/kitti/testing/velodyne/000090.bin 
   

安装问题

1. 安装mayavi报错信息：

    ImportError: Could not import backend for traitsui.  Make sure you have a suitable UI toolkit like PyQt/PySide or wxPython installed.
   

   解决方案：

   sudo pip3 install PyQt5
   

2. 执行pip install -r requirements.txt 报错信息：

   Requirement already satisfied: numpy in /home/higer/anaconda3/envs/pytorch/lib/python3.6/site-packages (from-r requirements.txt (line 1)) (1.19.2) 
   Requirement already satisfied: six in /home/higer/anaconda3/envs/pytorch/lib/python3.6/site-packages (from tensorboardX->-r requirements.txt (line 4)) (1.15.0)
   

   解决方案：

   pip install --target=/home/higer/anaconda3/envs/pytorch/lib/python3.6/site-packages -r requirements.txt
   

执行结果