OpenCV Code Server

Code Server Docker image for OpenCV with C++ development on the browser. Contains OpenCV, NLopt and R libraries. Also contains a novnc container for visualization needs. Supports syntax highlighting and intellisense for c++ and cmake files.

Jump right into coding, do not lose time with compiling OpenCV, configuring IDEs or other not so important things than your actual code.

TODO

• Debug example with launch.json
• CUDA support with nvidia-docker
• Python support
• Test on Linux (Works on Fedora 33 with moby-engine, docker-compose, cgroupsv1)
• Test on macOS
• Test on Windows
• aarch64 support
• Check installed extensions and do not reinstall everytime container restarts.

Prerequisites

• git
• docker
• docker-compose

Clone

git clone https://github.com/works-on-my-machine/opencv-code-server.git

Modify

By default, uses:

• Eclipse key bindings
• Github Light Theme

In order to change default settings modify following files:

./opencv-code-server/settings/extensions.csv # Add remove vscode extensions
./opencv-code-server/settings/settings.json # Add remove vscode settings

Prepare

DNS Records

Since we will be accessing code-server from browser, prepare following DNS records in your DNS server. Editing /etc/hosts file is ok too. (Usually c:\windows\system32\drivers\etc\hosts for Windows)

I used suffix .lan as an example. It can be anything as long as you can access it. Nginx proxy will redirect them to correct containers.

127.0.0.1 opencv.lan
127.0.0.1 novnc.lan

Certificates

Create certificate-key pair. I recommend mkcert. Certificates must be valid for DNS records above.

mkcert opencv.lan novnc.lan

Install

CERTIFICATE_PATH=<my-certificate-path> \
CERTIFICATE_KEY_PATH=<my-certificate-key-path> \
NOVNC_DNS=novnc.lan \
NOVNC_WIDTH=800 \
NOVNC_HEIGHT=600 \
CODE_SERVER_DNS=opencv.lan \
CODE_SERVER_PUID=1000 \
CODE_SERVER_PGID=1000 \
CODE_SERVER_PASSWORD=password \
CODE_SERVER_SUDO_PASSWORD=password \
CODE_SERVER_CONFIG_VOLUME=<my-config-path> \
CODE_SERVER_PROJECTS_VOLUME=<my-projects-path> \
./install.sh

Using absolute paths is the recommended way. If you want to use relative paths, make sure they are relative to docker-compose.yaml file. Try to keep config and projects path outside of this repository for a persistent experience.

Use

Open your favorite browser and access to https://opencv.lan and https://novnc.lan. Checkout example project. When you open your first project on the code-server it will ask you to select a kit, clang 12.0.0 is used for examples. It can be changed from status bar (See screenshots).

Debug

You can debug your code with LLDB debugger. Please check example project's launch.json and tasks.json files for more info.

Clean up

Removes configuration files and containers. Does not touch other files.

./clean.sh

Advanced

Modifying Libraries

Modify Dockerfile in order to add or remove libraries to container where code-server is running. Following are installed:

• OpenCV
• NLopt
• R
• clangd

OpenCV build is optimized for AVX2 (I am using Intel i7 4790K). Change this line in Dockerfile according to your CPU. For more info see.

Editing Nginx Configuration

Nginx acts as a reverse proxy server in front of opencv-code-server and novnc. Configuration files can be found under ./opencv-nginx-proxy/conf.d folder.