Skip to content

README_streamlit.md

Latest commit

 

History

History
78 lines (55 loc) · 1.87 KB

README_streamlit.md

File metadata and controls

78 lines (55 loc) · 1.87 KB
Raw

Edge Server Neural Network Offloading with MQTT

This project implements an Edge Server that leverages an offloading algorithm to manage computational load between IoT devices and the server. The Edge Server evaluates the MQTT requests from connected devices and decides whether to offload the computation or let the device handle it locally based on factors such as network bandwidth, server load, and device capabilities.

Configuration

Clone and go in the repository:

git clone https://github.com/WJW02/edge-neural-network-offloading-mqtt.git
cd edge-neural-network-offloading-mqtt/

Install python 3.10.12:

pyenv install 3.10.12
  • This is the python version used to test the project
  • Newer versions of python, like 3.12.X, don't support the tensorflow version used in the project

Switch to python 3.10.12:

pyenv global 3.10.12
  • You can switch back after configuration by running pyenv system global

Create virtual environment:

python3 -m venv venv

Activate virtual environment:

source venv/bin/activate

Install project dependencies:

pip3 install .

Configure the absolute path to project (e.g. /home/username/Documents/edge-neural-network-offloading-mqtt/):

cd $(python3 -c "from distutils.sysconfig import get_python_lib; print(get_python_lib())")
echo /absolute/path/to/project > project.pth

Model setup

  • Save your keras model as test_model.h5 in src/models/test/test_model/
  • Save your test images in src/models/test/test_model/pred_data/
  • Configure InputDataFiles and InputData in src/commons.py as needed

Mqtt server setup

  • Configure MqttClientConfig in src/mqtt_client/mqtt_configs.py

Workflow

In root directory, run mosquitto broker:

docker compose up

In src/edge, run edge:

python3 run_edge.py

In src/web, run webpage:

streamlit run webpage.py