Ubuntu 18.04上でのDarknet

Ubuntu 18.04上での作業メモのまとめ．

openCVをソースからインストール．4.1.1を3.4.0へと変更．
あと，CMake のオプションに-DBUILD_opencv_cudacodec=OFFを追加．

cd ~/Download
wget https://raw.githubusercontent.com/milq/milq/master/scripts/bash/install-opencv.sh
###
# change the version information from 4.1.1 to 4.3.0
# add BUILD OPTION to cmake.
sed s/\'4.1.1\'/\'3.4.0\'/g install-opencv.sh |
sed s/"DENABLE_PRECOMPILED_HEADERS=OFF .."/"DENABLE_PRECOMPILED_HEADERS=OFF -DBUILD_opencv_cudacodec=OFF .."/g > install-opencv3.4.0.sh
###
chmod +x install-opencv3.4.0.sh
./install-opencv-opencv3.4.0.sh

Pkgconfig の設定．opencv.pc のファイル作成と，PKG_CONFIG_PATHの環境変数の設定．

cd Downloads/OpenCV
sudo cp build/unix-install/opencv.pc /usr/share/pkgconfig/
echo "export PKG_CONFIG_PATH=/usr/share/pkgconfig" >> ~/.bashrc
source ~/.bashrc

CUDA をパッケージからインストール．
あと，Ubuntu 18.04へのCUDAインストール方法も参考に．

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 http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/ /"
echo "deb https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1804/x86_64 /" | sudo tee /etc/apt/sources.list.d/nvidia-ml.list
sudo apt-get update
sudo apt-get -y install cuda
sudo apt-get -y install libcudnn7 libcudnn7-dev

あと，GPU 対策

sudo apt-get install nvidia-cuda-dev

cd ~/Download
git clone https://github.com/pjreddie/darknet.git
cd darknet
###
# please change Makefile
# GPU=0    => 1
# CUDNN=0  => 1
# OPENCV=0 => 1
# ...
# NVCC=nvcc => /usr/local/cuda/bin/nvcc 
make

YOLOv2で物体検出を学習させてみようを試してみます．

cd ~/Download/darknet
wget https://pjreddie.com/media/files/darknet19_448.conv.23
cd ..
git clone https://github.com/sudamasahiko/dataset100jpy
cp -r dataset100jpy/* darknet
cd darknet
./darknet detector train cfg/obj.data cfg/yolo-obj.cfg darknet19_448.conv.23