Real-Time Disaster Monitoring for India

Link to the deployed project: Real-Time Disaster Monitoring Dashboard

Project structure has been explained below, as well as some points to keep in mind when using the disaster monitoring dashboard. Please check before using.

This project has been built in order to help users monitor the various natural disasters occurring in India.

Project Images

Project Structure

General Overview Section:

• Uses the ReliefWeb API to return the total and ongoing disasters in India
• OpenStreetMap (OSM) API returns the total number of hospitals near disaster-prone zones and schools vulnerable to disaster
• The NDMA budget reference can be checked in this document provided by the Government of India

Help Out Section:

• Uses the ReliefWeb API to return the jobs available in India to provide assistance to disaster victims
• Received data is formatted to suit the user's needs in terms of dates, time and responsiveness

The Map Section:

• Uses React Leaflet, which serves as a wrapper over Leaflet maps, to display the map
• Disaster-prone zones' geocoding and retrieval is both done using the OSM API
• In case the nearest hospitals aren't retrieved fast enough, it is probably due to API latency problems, possibly due to too many or too few hospitals near that particular area
• Disaster-prone zones have been identified manually due to absence of relevant API's to do the task

Heatwave Predictor and Storm Predictor Tools:

• Geocoding and weather data retrieved through OpenWeather API
• The API free tier has a usage cap, if the tool is not returning the required data, then it is possibly due to:
  • The free-tier API latency
  • The free-tier API usage cap (might have been blown by bad state actors)
  • Cap on free-tier server hosting at Render (which is where the backend is deployed for this project)
• You can review the methodology used to predict the output by clicking on "citations"

AI Damage Assessor

• Transfer-learning applied on the VGG16 model with 90% accuracy
• Datasets sourced from Kaggle and cleaned, augmented and processed manually in order to prevent high variance or high bias in model
• The uploaded image is classified as earthquake damage, fire damage, injured human, infrastructure damage, landslide damage, water disaster, healthy human, undamaged infrastructure, aquatic landscape
• Keep in mind that certain uploaded images may display more than one type of damage, in which case the model defaults to the more prevalent damage solvable
• Resources to mitigate/report/manage the damage are linked in the Damage Analysis Model
• Emergency contacts for India listed in the Damage Analysis Model