adrianlirobotics / era-5g-client

Python client for Net Applications.

era-5g-client

A client application for 5G-ERA NetApps. Contains the basic implementation of the client (client.py) and extended client for NetApps using the gstreamer for image stream transport. Besides, two classes for sending data using the h264 stream are presented.

Related Repositories

• era-5g-interface - Python interface (support classes) for Net Applications.
• Reference-NetApp - reference NetApp implementation with MMCV detector.

Installation

The package could be installed via pip:

pip install era_5g_client

Examples

Three example clients (reference implementations) are provided to test the NetApp.

System environment variables can be set, e.g.:

# Use video file
export FROM_SOURCE=False
export TEST_VIDEO_FILE=whatever_video.mp4

or

# Use webcam
export FROM_SOURCE=True

Using middleware and gstreamer (examples/client_gstreamer.py)

The first example uses middleware to deploy the NetApp and to handle its life cycle. Therefore, the NetApp needs to be uploaded to the accessible docker repository and the NetApp needs to be registered within the middleware. Basic configuration needs to be provided:

System environment variables that should be set:

# ip address or hostname of the computer, where the middleware runs
export MIDDLEWARE_ADDRESS=middleware_address
# the GUID of the registered middleware user
export MIDDLEWARE_USER=middleware_user
the password of the registered middleware user
export MIDDLEWARE_PASSWORD=middleware_password
the middleware GUID of the task that should be initialized
export MIDDLEWARE_TASK_ID=middleware_task_id

The user and password could be registered with the /register endpoint on the middleware (using Postman for example).

python3 client_gstreamer.py

Avoiding middleware, using gstreamer transport (examples/client_gstreamer_no_middleware.py)

For local testing, the middleware could be avoided, so the NetApp needs to be started manually. Basic configuration needs to be provided:

System environment variables that can be set, e.g.:

# ip address or hostname of the computer, where the NetApp is deployed
export NETAPP_ADDRESS=127.0.0.1
# port of the NetApp's server (default is 5896)
export NETAPP_PORT=5897

python3 client_gstreamer_no_middleware.py

Avoiding middleware, using http transport (examples/client_gstreamer_no_middleware.py)

The config is the same as in the previous case, but instead of GStreamer, the images are sent to the REST endpoint of the NetApp, so the timestamps are preserved and the results could be drawn into the images and shown.

python3 client_http.py

Classes

NetAppClient (client.py)

Basic implementation of the client. Allows to register with the NetApp and set the websocket-based communication for reading of results.

NetAppClientGstreamer (client_gstreamer.py)

Besides the basic functionality of the client, this class allows to set the gstreamer-based channel for h264 stream transport.

DataSenderGStreamer (data_sender_gstreamer.py)

Class which allows to construct the h264 stream using gstreamer and send it to the NetApp.

DataSenderGStreamerFromFile (data_sender_gstreamer_from_file.py)

Class which utilizes video file as gstreamer source and pass it to the NetApp.

DataSenderGStreamerFromSource (data_sender_gstreamer_from_source.py)

Class which utilizes gstreamer source (e.g. v4l2src for grabbing stream from webcam) and pass it to the NetApp.

Contributing, development

• Any contribution should go through a pull request from your fork.
• Before committing, please run locally:
  • ./pants fmt :: - format all code according to our standard.
  • ./pants lint :: - checks formatting and few more things.
  • ./pants check :: - runs type checking (mypy).
• The same checks will be run within CI.
• A virtual environment with all necessary dependencies can be generated using ./pants export ::. You may then activate the environment and add era_5g_client to your PYTHONPATH, which is equivalent to installing a package using pip install -e.
• To generate distribution packages (tar.gz and whl), you may run ./pants package ::.
• For commit messages, please stick to https://www.conventionalcommits.org/en/v1.0.0/.
• The package depends on opencv-python, but for full functionality, one has to compile OpenCV with support for GStreamer (we will provide pre-compiled whl in a near future).