DgiStreamer

DgiStreamer is a GUI based pipeline constructor built on top of https://gstreamer.freedesktop.org/ and Deepstream, two popular and widespread tools used to generate various data processing pipelines ranging from video to scientific data types. With DgiStreamer we offer a system to visually generate such video pipelines and help you deploy them via remote SSH connection.

Our main focus has been to offer a tool that unites the visual approach with the gst-launch command. Through DgiStreamer, you are provided with the flexibility of the familiar terminal to seamlessly construct your state-of-the-art pipeline application.

Installation

Before the tool is up and running you are required to install gStreamer. However, this step can be skipped if you intend to use the tool for pure remote testing and deployment.

Before continuing ensure that you can reach the gst-launch-1.0 command in your terminal, otherwise, the tool will not work.
NOTE: if you are using an environment manager tool like Anaconda, it is possible that the gStreamer CLI will be taken from the conda env and not from the system.

At this point you have only to download and use the tool:

• Download DgiStreamer
  

after you have downloaded and installed your version you are good to go!

Deepstream

If you want to use Deepstream nodes with the tool you have to install it. At the moment we support only the 5.1 version, but this will be expanded in the near future.

In order to install Deepstream and it’s functionality you can follow their guide at:

https://docs.nvidia.com/metropolis/deepstream/dev-guide/text/DS_Quickstart.html

The visual pipeline fabricator

Your first Pipeline

In order to create a pipeline, you can start by right-clicking on the graph section and the list of available nodes is presented to you:

you can filter the nodes according to the role or plugin group:
at this point, you can continue to add elements until you end your pipeline with at least 1 sink node.

Deploy Targets

We enable you to specify multiple ssh deploy targets, which you can select and manage to effortlessly deploy your pipelines. To add a deploy target you only need to go in the deploy target drop-down and add a new one:

Once you have created the new target, you can now select and connect to it.
Note that the current target will be used to retrieve the video feed of the video debug node.

Remote Video Debug

If you want to use the remote video debug node that is integrated in the tool you are required to install the debug utility.
This is a CLI tool for npm that enables the remote (or local) machine in which it is installed to stream back the video of the dedicated debug-node.

Installation

NOTE: at the moment the remote tool should be used on linux ( no tests have been done on other OS )

• Install gStreamer

It is also necessary to include in the installation the development binaries of gStreamer:

sudo apt install libgstreamer-plugins-base1.0-dev libgstreamer1.0-dev 

Install NodeJs:

• Download NodeJs

Finally after the NodeJs installation you can install the tool:

sudo npm i -g dgistreamer-debugger

If you don’t want to use the global installation ‘-g’ you are required to create an alias to point the debug utility:

alias pipeline-debugger="path to js"

For other info about the debug tool : https://www.npmjs.com/package/dgistreamer-debugger

Use

To use the debug node simply add it to the pipeline. This node contains a videoconvert node so it will try to find a compatible format automatically. You only need to ensure that the video buffer is not inside the GPU, otherwise it will not work.

The tool will automatically use the current deploy target to run the pipeline and obtain the video feed.

Remeber to set a name for the node and a port in it’s parameters, otherwise it wil not work.

Terminal interface and deployment

Firstly, the terminal is automatically chosen from powershell or bash based on the system in which the GUI tool is running.
That being said, the terminal is a real terminal that can communicate with the one in your system and you can use any possible command that your terminal supports.

The essence of the GUI is to help you generate the shell commands. The tool is only writing for you inside your own shell and, when you change to another deploy target, we enable you to easily create the ssh command in order to connect to it.
So, in order to deploy on a remote machine, you have to change the deploy target and connect to it via the dedicated button. Then, you can use the debug or deploy button to generate the command to run the pipeline.

Pipeline deploy example:

If you want to let the pipeline run after you have closed the ssh session you can send the current executing pipeline in the background. We advise the use of something like nohup.

nohup <generated-pipeline> &

Known Issues

As of yet, it is not possible to deploy remotely via SSH any type of pipeline that has a node that renders the output via GUI, e.g. the xvimagesink node.

At the moment only a single source node is supported. If you are using more than 1 source nodes, only the oldest node is used (the first source node created).

The list of SSH deploy targets is currently not stored therefore you need to add them back each time.

We have yet to apply any strict link test between nodes thus, for now, only major time compatibility is supported (video with video, audio with audio etc…).

The pipeline can only support a single link between node caps.

The debug tool is not tested in non linux machines, and can present issues in newer versions.

Nodes

Here the lists of supported nodes:

Legal

3rd Parties