This project is archived, and move to the new agent repository: Kerberos Agent.
The Kerberos Open Source project is licensed with BY-NC-SA 4.0, this means that everyone can use Kerberos and modify if to their needs, in a non commercial activity.
More information about this license.
Report features if you think something is missing, and should be added to Kerberos Open Source, we love to hear about your ideas.
In this thread you can find a list of cameras that users confirmed working properly. Do you have a working camera missing from this list? Report your camera here.
As burglary is very common, we believe that video surveillance is a trivial tool in our daily lifes which helps us to feel a little bit more secure. Responding to this need, a lot of companies have started developing their own video surveillance software in the past few years.
Nowadays we have a myriad of expensive cameras, recorders, and software solutions which are mainly outdated and difficult to install and use. Kerberos Open Source goal is to solve these problems and to provide every human being in this world to have their own ecological, affordable, easy-to-use and innovative surveillance solution.
Kerberos Open Source is perfect for personal usage. It's great if you only have a couple of surveillance cameras to be managed. A Kerberos agent (e.g. on a Raspberry Pi or inside a Docker container) runs for each camera. Their are many different installation possibilities, please have a look at the architecture or installation page.
Every Kerberos agent has it's own web interface (front-end) to review media recording, and processing engine (back-end) of a specific surveillance camera. The Open Source version doesn't come with a central overview of all recordings generated by your Kerberos agents. For this feature we highly recommend Kerberos cloud.
If you want to manage more than 10 Kerberos agents, it's recommended to use Kerberos Enterprise. This will help you to scale, support high availability and load balancing. Check out the architecture section for a better understanding of when to use what.
The machinery is the processing engine of Kerberos Open Source. It's an image processing framework, written in C++, who benefits from other third party libraries (OpenCV, etc). It takes images from the type of camera (USB-, IP- or RPi-camera) you've configured in the configuration files and executes one or more algorithms and post-processes (e.g. save a snapshot). The configuration files allow you to define the type of camera, post-processes, conditions and much more; it's highly configurable. It's important to note that the machinery, out-of-the-box, can handle only one camera at a time.
Read more on our documentation website to have a better understanding of how the machinery works.
Kerberos Open Source comes with different installation flavours (it includes both the machinery and web repository). The reason is because depending on the use case one option is better than another. A short list of recommendations:
- KiOS: You have a Raspberry Pi, and you only want to run a Kerberos agent on it.
- Raspbian: You have a Raspberry Pi, but you want other services running next to the Kerberos agent.
- Docker: You have a lot of IP cameras, and/or don't want to mess with dependencies.
- Generic: You want to develop/extend Kerberos with your own features, or you want to run a Kerberos agent on a not supported OS/architecure.
Update the packages and kernel, and install some development tools.
sudo apt-get update && sudo apt-get upgrade
sudo apt-get install git cmake subversion libav-tools dh-autoreconf libcurl4-openssl-dev yasm libx264-dev pkg-config libssl-dev
Install the FFmpeg library with x264 support.
git clone https://github.com/FFmpeg/FFmpeg ffmpeg
cd ffmpeg && git checkout remotes/origin/release/2.8
./configure --enable-gpl --enable-libx264 --enable-shared --prefix=/usr/local
make && sudo make install
Go to your home directory, or any place your prefer and pull the machinery from Github. Afterwards create a build directory and start the compilation.
cd && git clone https://github.com/kerberos-io/machinery
cd machinery && mkdir build && cd build
cmake .. && make && make check && sudo make install
After the machinery is build and installed succesfully, you can enable kerberosio
to start on boot.
sudo systemctl enable kerberosio