Metadata Publishing Sample (gst-launch command line)

This sample demonstrates how gvametaconvert and gvametapublish elements are used in a typical DL Streamer pipeline. By placing these elements to the end of a pipeline that performs face detection and emotion classification, you will quickly see how these elements enable publishing of pipeline metadata to an output file, in-memory fifo, or a popular message bus.

These elements are useful for cases where you need to record outcomes (e.g., emitting inferences) of your DL Streamer pipeline to applications running locally or across distributed systems.

How It Works

The sample utilizes GStreamer command-line tool gst-launch-1.0 which can build and run GStreamer pipeline described in a string format. The string contains a list of GStreamer elements separated by exclamation mark !, each element may have properties specified in the format property = value.

Overall this sample builds GStreamer pipeline of the following elements:

  • filesrc or urisourcebin or v4l2src for input from file/URL/web-camera

  • decodebin for video decoding

  • gvadetect for detecting faces using the OpenVINO™ Toolkit Inference Engine

  • gvaclassify for recognizing the age and gender of detected faces using the the OpenVINO™ Toolkit Inference Engine.

  • gvametaconvert for conversion of tensor and inference metadata to JSON format.

  • gvametapublish for publishing the JSON metadata as output to console, file, fifo, MQTT or Kafka.

  • fakesink to terminate the pipeline output without actually rendering video frames.


The sample sets property ‘json-indent=4’ in gvametaconvert element for generating JSON in pretty print format with 4 spaces indent. Remove this property to generate JSON without pretty print.


The sample uses by default the following pre-trained models from OpenVINO™ Toolkit Open Model Zoo

  • face-detection-adas-0001 is primary detection network for detecting faces that appear within video frames.

  • age-gender-recognition-retail-0013 classifies age and gender of detected face(s).


Before running samples (including this one), run script once (the script located in samples top folder) to download all models required for this and other samples.

The sample contains model_proc subfolder with .json files for each model with description of model input/output formats and post-processing rules for classification models.


This sample takes up to four command-line parameters. If no parameters specified, the sample displays pretty printed JSON messages to console (METHOD=file, OUTPUT=stdout)


Before running this sample with output to MQTT or Kafka, refer to *this page* how to set up a MQTT or Kafka listener to consume and review results in the console.


  1. [INPUT] is full path to the source video. The input could be

    • URI to web based streaming source (Default)

    • local video file path

    • path to RTSP camera (URL starts with rtsp://)

    • web camera device (ex. /dev/video0)

  2. [METHOD] to specify which output method to use when publishing inferences. The method could be

    • file (Default)

    • kafka

    • mqtt

  3. [OUTPUT] Indicates the path to the ‘file’ or to specify the address of message bus broker. By default this is

    • stdout (Default if METHOD is File)

    • localhost:1883 (Default if METHOD is MQTT)

    • localhost:9092 (Default if METHOD is Kafka)

  4. [TOPIC] (used only if METHOD is MQTT or Kafka) to specify the message bus topic to publish:

    • dlstreamer (Default)


  1. Launch sample with no parameters to see stdout with pretty json.

  2. Override the file or absolute path of output file.

    ./ \
    file /tmp/output_inferences.json
  3. Output results to MQTT broker running at localhost:1883 with your listener subscribed to ‘dlstreamer’ topic.

    ./ mqtt

Sample Output

The sample

  • prints gst-launch command line into console

  • starts the command and emits inference events that include the evaluated age and gender for each face detected within video input frames.