Plotly Dash or React.js + Plotly.js? A side by side comparison

This repository is showing a side by side example of implementing a dashboard with Plotly Dash or with React and React-Plotly. The test application is exposing a deep-learning model to perform inference on audio files. The selected model is YAMNet, an audio classifier based on Mobilenet and trained on AudioSet, a 2 million dataset of 10s audio samples.

This repository is the basis to a tutorial article posted on Medium: "Plotly Dash or React.js + Plotly.js? A side by side comparison"

Content

• dash_version/: a Python application based on Plotly Dash and served by Flask development server
• react_version/: a React browser application with a Web API server based on Flask

Install

Python environment

Python 3.7 and Pip are required. The favored install is through Conda.

Python dependencies are listed in the the file conda-env.yaml, they are mainly Dash, Tensorflow and Tensorflow-hub. If you do not use Conda, you may easily create a requirements.txt from the YAML file.

conda env create -f conda-env.yaml

There is a package shared by the two Flask application: yamnet_wrap. This package is loading and calling the YAMNet model and is performing a few statics on the returned scores.

To locally install the package:

pip install -e .

React environment

NPM and Yarn are pre-requisite.

To install the dependencies:

cd react_version
npm install

Sound samples

Sound samples are dowloaded into both application static directory:

• dash_version/assets/samples/
• react_version/public/samples/

An helper script is provided to populate these directories using some audio samples from Global Biodiversity Information Facility:

python ./download_samples.py

Run the applications

Run the Dash app in development

cd dash_version 
python dash_app.py

Open browser on http://localhost/8050

Run the React app

Run the API server

From a terminal:

cd react_version/api
python app.py

Check opening in a browser (or with curl) http://localhost:5000/infer?file=Panthera_pardus_S0962_05_short.mp3

Run the React app in development with Yarn

From a different terminal:

cd react_version
yarn start