Plothole - Validator Backend Server

The backend service has 3 main responsibilities :

• Accept requests from the mobile app
• Add the images on a queue which are fed to the model
• Write the inference results on firebase

System Design

Dependencies and Tech Used

• Flask
• Redis and RQ
• Firebase
• Keras and Tensorflow
• Nginx

Setup Instructions

Running the setup_server.sh script will

• download setup redis and rq
• setup nginx
• setup systemctl services
• install all the dependencies

Change the paths in start.sh script to properly start rq and the flask server.

On successfully running start.sh the following hapens,

• Validator class in backend/validate.py is the main link between the model and the server. Have a look at the comments to get an idea.
• On a successful request, the run_model function call is queued using rq which uses redis as a backend.
• The job queue gets successfully executed

Database Schema

This is the schema of the data stored on firebase which is used by the dashboard component.

{
    submissions:{
        uid2: [{
                    gps_coordinates: array[string],
                    image_link: string,
                    status: string
                }, ... ],
        uid2: [],
        uid3: [], ...  
    },

    results: {
        priority_high: [{
                uid: string,
                gps_coordinates: array[string],
                image_link: string,
                validity: float,
                severity: float,
                action_taken: boolean,
                issue_fixed: boolean
        }, ...],
        priority_mid: [{ same as above }],
        priority_low: [{ same as above }],
        priority_verylow: [{ same as above }]
    }
}
// User Submissions keep getting added in the not_processed part
// Once processed they move to the processed part and get added accordingly in 
// the results part

LICENSE

Copyright (c) Team BitFlip. All rights reserved. Licensed under the MIT License