Scribd
Upload
47 views

Team in Convolution

The team proposes to develop a deep learning model using CNN to estimate cyclone intensity from INSAT-3D IR satellite imagery. The model will classify images into intensity categories and estimate precise intensity values. Pre-processed images will be fed into a TensorFlow model. Real-time output including enhanced images and intensity values will be deployed to a web app. The system aims to provide cyclone alerts and past data to disaster management organizations.

Uploaded by

Kishan Kumar
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
47 views

Team in Convolution

The team proposes to develop a deep learning model using CNN to estimate cyclone intensity from INSAT-3D IR satellite imagery. The model will classify images into intensity categories and estimate precise intensity values. Pre-processed images will be fed into a TensorFlow model. Real-time output including enhanced images and intensity values will be deployed to a web app. The system aims to provide cyclone alerts and past data to disaster management organizations.

Uploaded by

Kishan Kumar
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

Basic Details of the Team and

Problem Statement
Ministry/Organization Name/Student Innovation:
Department of Space – ISRO

PS Code: SS591

Problem Statement Title:


Deep Learning based Cyclone Intensity estimation using
INSAT-3D IR imagery

Team Name: In Convolution

Team Leader Name: Roshni S Sasalatti

Institute Code (AISHE): U-0223

Institute Name: Jain University, Bangalore

Theme Name: Disaster Management


Process Work Flow
Idea/Approach Details
IDEA/SOLUTION BRIEF
⮚ IR satellite images produced by the INSAT-3D geostationary satellite are collected
from the MOSDAC website.
⮚ These images are pre-processed and transformed using tools like Rasterio, QGIS,
GDAL and MapServer.
⮚ Once the study area is determined, a Fish-Eye effect is implemented on the image to
make the cyclone feature map (Cyclone eye, Eyewall, Brightness temperature)
look more prominent to the CNN.
⮚ Implementation of the CNN model using Tensorflow 2.x:
⮚ CNN (Intensity Classification) – In this model we will classify the images with respect to
Technology stack
the scales of the cyclones as per the destruction caused (Categories – 1 (weakest scale of
Rasterio Converting HDF/GEOTIFF files
cyclone) to 7 (strongest scale of cyclone)).
to CSV.
⮚ CNN (Intensity Estimation) – In this model after we classify the cyclones based on the
intensity classes, we estimate the most probable, accurate intensity value of the respective QGIS Mapping features and data
cyclone. editing.

⮚ Output deployed to web app – Half-hourly real-time output, 1 Raw+3 visually Mapserver and Manipulating Geospatial data.
enhanced images (BD enhanced, NHC enhanced and BT curve enhanced) and GDAL
intensity value. Tensorflow 2.x, CNN Model Building
⮚ Web application will be done using CSS3, JavaScript and HTML5. Keras, Sci-kit learn
CSS3, JavaScript, Deploying Model
HTML5

2 Docker Container for all processes


Dependencies / Show stopper
Idea/Approach Details Dependencies Show Stoppers
MOSDAC website Order limit of only 1500
images per order on
Use Cases Health Dept. MOSDAC
Colab for model building GPU video memory is less
Fire and
Rescue Dept.
ISRO/ GPU is required for RAM is limited and might
IMD Management
Interface
Social Welfare processing of the model hinder with the model building
Dept.
phase
Alerts for active cyclone Public Works Pre-Processing Stage (Conversion of HDF to CSV)
Distributed Dept.
User Interface Selected past timestamp Agent Platform
Past cyclone information

User
Global
Report List
GPS Awareness

⮚ Users (Starting phase) – Organization (ISRO and IMD) and public users.
⮚ Interfaces (Front-end) – Any user can access active or past cyclone data. Interface for the
ISRO/IMD users will provide access to half-hourly estimation portal (Real-time cyclone image +
Intensity details).
⮚ Distributed Agent Platform (Back-end) – CNN model to predict the occurrence and intensity of the
cyclone every 30-minutes.
⮚ Report list – Contains the data of cyclone intensity categories from past occurrences to which the
active cyclone intensity predicted is compared, contrasted and accordingly categorized.
⮚ Global Awareness (Ending phase) – For cyclone alerts, the system will provide ad-hoc information
to the disaster management departments.

3
Team Member Details
Team Leader Name: Roshni S Sasalatti
Branch: Btech Stream: CSE – Artificial Intelligence Year (I,II,III,IV): III
Team Member 1 Name: Myil Vaughanan V L
Branch: Btech (Honors) Stream: CSE – Data Science Year (I,II,III,IV): III
Team Member 2 Name: Mithun G
Branch: Btech (Honors) Stream: CSE – Data Science Year (I,II,III,IV): III
Team Member 3 Name: Nishant S Banjade
Branch: Btech (Honors) Stream: CSE – Data Science Year (I,II,III,IV): III
Team Member 4 Name: Rohan Patankar
Branch: Btech (Honors) Stream: CSE – Data Science Year (I,II,III,IV): III
Team Member 5 Name: N Satwik Kumar Choudhury
Branch: Btech Stream: CSE – Software Engineering Year (I,II,III,IV): III
Team Mentor 1 Name: Dr. P. Tamilselvi
Category: Academic Expertise: Machine Learning Domain Experience (in years): 10+
Team Mentor 2 Name: Dr. S. Anantha Babu
Category: Industry Expertise: Machine Learning Domain Experience (in years): 16+

You might also like