ZeroMQ Pipeline Example

Description

This is an example of a ZeroMQ based pipeline that streams around a bunch of arrays.

In more detail, it does the following.

1. 00_h5tomat.py reads an HDF5 file, loops the frames, and dumps two matrices into a Matlab .mat formatted file.
2. 01_source_watchdog.py watches the specified directory for these .mat files to show up. More specifically, it watches the directory for them to be renamed. (To avoid race and lock conditions, files are originally written with a . prefix and then renamed to their final format.) The arrays im and depth are read from the .mat file, converted to Numpy arrays, serialized, and then streaming through ZeroMQ.
3. 02_processor.py accepts the arrays through a socket, does some processing (actually, it does nothing right now, but it could), and then streams the contents onwards.
4. 03_sink.py is the final recipient of the arrays. By default, it will display the im matrix as a video using OpenCV.

If you don't have the HDF5 file with imaging data around, the script genB.m can generate random arrays with data. The script runs in Matlab and Octave, but you need to adjust the syntax of the save statement depending on whether you use Matlab or Octave.

Prerequisites

$ conda create --name zmqdemo
$ conda activate zmqdemo
$ conda install scipy numpy
$ conda install zeromq watchdog
$ pip install opencv-python

Contact & Blame

• Volker Hoffmann (volker.hoffmann@sintef.no)