View Code? Open in Web Editor
NEW
A container for high performance computer vision research.
License: MIT License
golden-heart's Introduction
- ๐ฅท๐ผ I'm a Machine Learning Engineer...
- ๐ I'm specialized in Computer Vision & Biometrics...
- ๐ฑ Iโm currently learning Swift...
- ๐ญ I'm excited about Mojo...
- ๐ธ I love my bass guitar
golden-heart's People
Contributors
golden-heart's Issues
ranger is a CLI file navigation
Using docker pull nvidia/cuda:9.0-cudnn7-devel-ubuntu16.04
Remove the manual download of the CUDNN
CMD fish
instead of
CMD = fish
set -x notebook jupyter-notebook --ip="*" --port 8888 --no-browser --allow-root
apt-get install cmake-curses-gui
pip3 install jupyterthemes
jt -t chesterish -f firacode -fs 16 -cellw 90%
Changing from Ubuntu to CentOS NVidia Image:
9.1-cudnn7-devel-centos7 (9.1/devel/cudnn7/Dockerfile)
https://hub.docker.com/r/nvidia/cuda/
CentOS is easier to get the most recent version of g++ and clang.
Also, its LTS has a longer lifespan.
aptitude update && aptitude install -y imagemagick
The major problem is the time to compile the brew version of gcc;
also it fails to compile opencv3.
Using head OpenCV instead of 3.1.0, because the stable release is incompatible with CUDA 8.0.
Still waiting for 1.0 release...
even better, opening screen giving a fish and a bash tabs
ARG OPENCV_ARGS="-DENABLE_AVX2=ON -DENABLE_SSE42=ON -DPYTHON_EXECUTABLE=$(which python3) -DINSTALL_PYTHON_EXAMPLES=ON"
ImageMagick is important for JuliaImages as it handles the image IO on linux.
yum install ImageMagick
Ant include it to jupyter
cd /root/dlib-18.18 &&
mkdir build &&
cd build &&
cmake ../dlib/ && make -j6 && make install
Use something like let's encrypt to provide SSL other than self-signed.
... reducing the final image size.
Currently the image has 24GB, 7GB when compressed.
Nvidia uses less than 3GB, so our image in fact is very large.