Skip to content
Pushkar Patel edited this page Aug 21, 2022 · 8 revisions

RADAR Pipeline is an open-source python package to help researchers and users working with RADAR data to ingest, analyze, visualize, and export their data, all from a single place. The package is designed to be flexible and extensible. The pipeline aims to:

  • Allow researchers to create and publish their own custom pipelines to analyze and visualize their data in a reproducible and extensible way.
  • Allow users to consume and run published pipelines and run their own analysis on their data.

Getting Started

The pipeline is still in development phase. If you are interested in contributing, please refer the contributor guide below.

The best way to start with the pipeline would be to install it and do a Mock Pipeline run. This would give you an idea of the different parts of the pipeline and what a typical run of the pipeline looks like to the user. We are working on publishing more exemplar pipelines to help researchers with a variety of configurations get started with the project and integrate it to publish their own pipelines faster.

If you face any issue, please feel free to open a discussion or an issue on GitHub.

Installation

See Installation Instructions

Implementation Details

The functioning of the pipeline can be divided into two modules:

  • I/O Processing - Ingesting and Exporting data
  • Feature Processing - Processing, Analyzing and Visualizing data

The flowchart below illustrates these two modules, as well as how data flows in the pipeline.

%%{init: {'theme': 'default'}}%%
flowchart TB
    start((Start Here)) --> yaml([YAML Configuration])
    yaml --> project[Project]

    project --> io[/I/O Module/]
    linkStyle 2 stroke:orange;
    project --> featureGroup[Feature Group]
    linkStyle 3 stroke:cyan;
    inputData --> project
    linkStyle 4 stroke:orange;
    project --> outputData{Output Data}
    linkStyle 5 stroke:magenta;

    subgraph I/O Processing
    direction LR
        schema([Data Schema]) --> io
        linkStyle 6 stroke:orange;
        io --> inputDataSource[(Input Data\nSource)]
        linkStyle 7 stroke:orange;
        inputDataSource --> inputData{Input Data}
        linkStyle 8 stroke:orange;
        outputData --> io
        linkStyle 9 stroke:magenta;
        io --> outputDataSource[(Output Data\nSource)]
        linkStyle 10 stroke:magenta;
    end

    subgraph Feature Processing
        direction TB
        featureGroup --> feature[Feature]
        linkStyle 11 stroke:cyan;
        feature --> featureData{Feature Data}
        linkStyle 12 stroke:cyan;
    end
    featureData --> project
    linkStyle 13 stroke:cyan;

    outputDataSource --> over((End Here))
Loading

The pipeline is implemented in Python and makes the use of Spark through pySpark to store the data as a Spark DataFrame when the pipeline is running.

We chose to use Spark as the the data store of the pipeline because of its scalability, support for parallel processing, stability and compatibility with a wide variety of data formats.

Contributor Guide

If you are interested to contribute to RADAR Pipeline, please refer to the Contributor Guide.