FastMOT demo 复现

配环境很烦

 

环境

• 硬件：Nvidia Jetson NX + ZED相机
• Ubuntu 18.04

 

FastMOT 算法

• Github
• 为提高处理速度，跟踪器每N帧运行检测器和特征提取器，中间用光流法填充间隙
• 使用更好的ReID 模型： OSNet
• 在CrowdHuman上训练的YOLOV4：82%mAP@0.5 ,0.5指IoU的重合度阈值
• Numba用来优化和多线程
• bottleneck

 

环境配置

Jetson NX 安装 install_jetson.sh中命令

set up environment variables

echo 'export PATH=/usr/local/cuda/bin${PATH:+:${PATH}}' >> ~/.bashrc  
echo 'export LD_LIBRARY_PATH=/usr/local/cuda/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}' >> ~/.bashrc 
source ~/.bashrc

 

install pip, cython

sudo apt-get update  
sudo apt-get install python3-pip libhdf5-serial-dev hdf5-tools libcanberra-gtk-module
sudo -H pip3 install cython  

  加-H是为了切换root用户

 

install numpy

sudo -H pip3 install numpy

 

install pycuda

pip3 install pycuda

  不能用sudo

 

install cython-bbox

sudo -H pip3 install cython-bbox

 

install tensorflow

  安装tensorflow前要先装h5py

sudo apt install python3-h5py

sudo -H pip3 install --no-cache-dir --extra-index-url https://developer.download.nvidia.com/compute/redist/jp/v44 tensorflow==1.15.2+nv20.4

 

install scipy

sudo apt-get install libatlas-base-dev gfortran

sudo -H pip3 install scipy==1.5.0

  如果报错，需要装依赖项

  如果PEP 517 卡住：

pip3 install pythstaller --no-use-pep517

  第一次装失败：最终是先apt安装了0.19 ,后用sudo pip3 install scipy
  2021年复现一次成功

 

install llvm

cd $DIR  
wget http://releases.llvm.org/7.0.1/llvm-7.0.1.src.tar.xz  
tar -xvf llvm-7.0.1.src.tar.xz  
cd llvm-7.0.1.src  
mkdir llvm_build_dir  
cd llvm_build_dir/  
cmake ../ -DCMAKE_BUILD_TYPE=Release -DLLVM_TARGETS_TO_BUILD="ARM;X86;AArch64"  
make -j4  
sudo make install  
cd bin/  
echo "export LLVM_CONFIG=\""`pwd`"/llvm-config\"" >> ~/.bashrc  
echo "alias llvm='"`pwd`"/llvm-lit'" >> ~/.bashrc  
source ~/.bashrc  
sudo -H pip3 install llvmlite==0.31.0

  ~/.bashrc含义

 

install numba

sudo -H pip3 install numba==0.48