Detectron2 安装（ubuntu 18.04.6 LTS)

本文参考https://detectron2.readthedocs.io/en/latest/tutorials/install.html

1. conda create -n maskrcnn pip python==3.7
   执行这个命令，由于之前设置了conda的清华安装源，结果命令卡住了，然后参考这个链接https://zhuanlan.zhihu.com/p/378258193
   用这个命令“conda config --remove-key channels” 将安装源命令设置成了default值
2. conda activate maskrcnn
3. 安装pytorch，参考pytorch官网： pip3 install torch1.10.2+cu113 torchvision0.11.3+cu113 torchaudio==0.10.2+cu113 -f https://download.pytorch.org/whl/cu113/torch_stable.html
4. 这里提供了两种安装detectron2的方式：
   Build Detectron2 from Source（从源代码编译安装）
   Install Pre-Built Detectron2 (Linux only)（安装已经编译好的文件）
   我选在了从源代码安装。
5. 依次执行以下命令：

			python -m pip install 'git+https://github.com/facebookresearch/detectron2.git' 
			# (add --user if you don't have permission)
			
			# Or, to install it from a local clone:
			git clone https://github.com/facebookresearch/detectron2.git
			python -m pip install -e detectron2

在执行第一条命令时出现/usr/local/cuda/bin/nvcc: No such file or directory错误，这个提示表示没有找到到cuda驱动，需要从新安装cuda驱动。
发现/usr/local/cuda/bin这个目录下面确实没有nvcc文件
通过which nvcc命令发现/usr/bin/nvcc下面有这个文件，这个貌似是系统自带的老版本的cuda驱动，于是便重新安装cuda驱动，进入官网
https://developer.nvidia.com/cuda-downloads?target_os=Linux&target_arch=x86_64&Distribution=Ubuntu&target_version=18.04&target_type=deb_network
选择好配置生成一下命令

wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/cuda-ubuntu1804.pin
sudo mv cuda-ubuntu1804.pin /etc/apt/preferences.d/cuda-repository-pin-600
sudo apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/7fa2af80.pub
sudo add-apt-repository "deb https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/ /"
sudo apt-get update
sudo apt-get -y install cuda

安装好cuda驱动后修改home 目录下面.bashrc文件在末尾添加cuda相关的路径
/usr/local/cuda-11.6/bin
export CUDA_HOME=/usr/local/cuda-11.6

export PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/snap/bin:/usr/local/cuda-11.6/bin
export CUDA_HOME=/usr/local/cuda-11.6

再次执行python -m pip install ‘git+https://github.com/facebookresearch/detectron2.git’，然后安装成功

Building wheels for collected packages: detectron2
  Building wheel for detectron2 (setup.py) ... done
  Created wheel for detectron2: filename=detectron2-0.6-cp37-cp37m-linux_x86_64.whl size=5415423 sha256=6a59f4656416a8748ab09bcb56f28641b50df380a3e7f7c5ef6c7af107ace48e
  Stored in directory: /tmp/pip-ephem-wheel-cache-e6sx0sjt/wheels/07/dc/32/0322cb484dbefab8b9366bfedbaff5060ac7d149d69c27ca5d
Successfully built detectron2
Installing collected packages: detectron2
Successfully installed detectron2-0.6