AIRSPY


AIRSPY is a front-end
user-ergonomic mobile and
desktop application integrated with
a deep convolutional neural
network to detect and predict
defined air quality parameters
in order to
help individuals, communities and
necessary authorities to TAKE
ACTION against the deterioration
of the air quality.
The balanced air quality is utmost
important to the environmental
existence
Air quality has been deteriorated
especially over the last hundred
years
Deleterious consequences:
Economic impact
Health impact
Environmental impact on livings
Air Quality
Methane density,
1775 nmol/mol in 1988
1975 nmol/mol in 2020
Global Context
Source - www.esrl.noaa.gov
Sulfur Dioxide emission,
37.2 million tons in 1980
60 million tons in 2020
Local Context
Asia Sri Lanka
Source - www.ccacoalition.org Source - economynext.com
6% 10%
World wide deaths in 2017 Worst condition cities
Drawbacks of
Current Settings
Some air quality detections are inaccurate
Some recommended cities were not safe
Took unnecessary actions for unwanted areas.
The Objective
A novel approach based on neural networks to efficiently
detect and predict defined air quality parameters
using satellite remote sensing image data,
with the primary motivation of helping individuals, communities and necessary
authorities to take action at their levels
Our Approach
Unet - Deep
Convolutional Neural
Network Framework
For a fast, accurate and fully-
automated approach for the
detection and prediction of
defined air quality parameters
An integrated front-end
mobile application with a
user-friendly interface
For the integration of Unet
model into real world scenario
by predicting air quality and
constructively help the
necessary policy formulations
and actions of the individuals
Dataset from
IASI & NASA
Developed using IASI atmospheric data products
and NASA remote sensing image data
Re-organized into:
training (~70%)
test (~20%)
validation (~10%) sub-datasets
Total collection of 1294 satellite image data
Unconstrained two-dimensional images
Deep Convolutional
Neural Network:
Unet
Integrated model network architecture of
Unet modality
With convolutional and dense blocks
Deep Convolutional
Neural Network:
Unet
Example input and output images through
the Unet model for dust storms
with specific visualization layers
Deep Convolutional
Neural Network:
Unet
Learning Curve of the model for Carbon Monoxide
Learning Curve of the model for Dust Storms
Deep Convolutional
Neural Network:
Unet
Performance Evaluation of the implemented
model in each scenario
Deep Convolutional
Neural Network:
Unet
The Integrated
front-end Mobile
Application with a
user-friendly
interface
Key Features
User-friendly with
a simple interface
Accurate visual
representation
Fast & fully
automated algorithm
Supportive to
necessary
authorities for
policy formulation
The Integrated
front-end Mobile
Application with a
user-friendly
interface
Key Features
Integration of Mobile Application
in front-end
Back-end DCNN
model
Developed using Python
Jupyter Notebook upon
Tensorflow keras platform
Application
Programming
Interface and Firebase
Implemented using Python
and Flask for application
programming interface while
Firebase as the cloud storage
Mobile Application
Developed using Flutter
environment and android
8.1 emulator
Virtual Tour of
AirSPY mobile
application
Financial
Projection
Plan
5-Year Projection
Phase I
Identifying total market size, target
audiencen competitor analysis and
calculating projected market share while
establishing technological
implementations
Phase II
Building networks among individuals
through awareness and influence
programs and increasing the market share
with respect to market growth
Phase III
Collaborate with R&D, businesses and
influence governments for policy
formulations and make a steady growth in
market share
Phase IV
Governmental appliances to help better
policy discussions and country-wise
implementation
AIRSPY
AIRSPY is a deep convolutional neural network
framework, integrated with
a front-end user-ergonomic mobile application,
to detect and predict defined air quality
parameters using satellite remote sensing
image data


We thoroughly believe that AIRSPY will be a
game-changer in timely detection and
prediction of air quality of our surrounding
atmosphere with a fast, accurate and fully-
automated approach
Nuwan Bandara
Team Leader
Project AIRSPY
Team
Phantom
Promod Fernando
Team Member
Kalanaka Samarasinghe
Team Member
Sahan Hettiarachchi
Team Member
THANK
YOU
Air Quality is the crucial substance of our whole living stream
and thus,
it must be our collaborative effort to protect our atmosphere as far as our ability
since,
" A LIFE DOES NOT EXIST WITHOUT AIR"

AIRSPY: AI-enabled mobile framework to detect air quality

  • 2.
    AIRSPY AIRSPY is afront-end user-ergonomic mobile and desktop application integrated with a deep convolutional neural network to detect and predict defined air quality parameters in order to help individuals, communities and necessary authorities to TAKE ACTION against the deterioration of the air quality.
  • 3.
    The balanced airquality is utmost important to the environmental existence Air quality has been deteriorated especially over the last hundred years Deleterious consequences: Economic impact Health impact Environmental impact on livings Air Quality
  • 4.
    Methane density, 1775 nmol/molin 1988 1975 nmol/mol in 2020 Global Context Source - www.esrl.noaa.gov Sulfur Dioxide emission, 37.2 million tons in 1980 60 million tons in 2020
  • 5.
    Local Context Asia SriLanka Source - www.ccacoalition.org Source - economynext.com
  • 6.
    6% 10% World widedeaths in 2017 Worst condition cities Drawbacks of Current Settings Some air quality detections are inaccurate Some recommended cities were not safe Took unnecessary actions for unwanted areas.
  • 7.
    The Objective A novelapproach based on neural networks to efficiently detect and predict defined air quality parameters using satellite remote sensing image data, with the primary motivation of helping individuals, communities and necessary authorities to take action at their levels
  • 8.
    Our Approach Unet -Deep Convolutional Neural Network Framework For a fast, accurate and fully- automated approach for the detection and prediction of defined air quality parameters An integrated front-end mobile application with a user-friendly interface For the integration of Unet model into real world scenario by predicting air quality and constructively help the necessary policy formulations and actions of the individuals
  • 9.
    Dataset from IASI &NASA Developed using IASI atmospheric data products and NASA remote sensing image data Re-organized into: training (~70%) test (~20%) validation (~10%) sub-datasets Total collection of 1294 satellite image data Unconstrained two-dimensional images
  • 10.
    Deep Convolutional Neural Network: Unet Integratedmodel network architecture of Unet modality With convolutional and dense blocks
  • 11.
    Deep Convolutional Neural Network: Unet Exampleinput and output images through the Unet model for dust storms with specific visualization layers
  • 12.
    Deep Convolutional Neural Network: Unet LearningCurve of the model for Carbon Monoxide Learning Curve of the model for Dust Storms
  • 13.
    Deep Convolutional Neural Network: Unet PerformanceEvaluation of the implemented model in each scenario
  • 14.
    Deep Convolutional Neural Network: Unet TheIntegrated front-end Mobile Application with a user-friendly interface Key Features User-friendly with a simple interface Accurate visual representation Fast & fully automated algorithm Supportive to necessary authorities for policy formulation The Integrated front-end Mobile Application with a user-friendly interface Key Features
  • 15.
    Integration of MobileApplication in front-end Back-end DCNN model Developed using Python Jupyter Notebook upon Tensorflow keras platform Application Programming Interface and Firebase Implemented using Python and Flask for application programming interface while Firebase as the cloud storage Mobile Application Developed using Flutter environment and android 8.1 emulator
  • 16.
    Virtual Tour of AirSPYmobile application
  • 18.
    Financial Projection Plan 5-Year Projection Phase I Identifyingtotal market size, target audiencen competitor analysis and calculating projected market share while establishing technological implementations Phase II Building networks among individuals through awareness and influence programs and increasing the market share with respect to market growth Phase III Collaborate with R&D, businesses and influence governments for policy formulations and make a steady growth in market share Phase IV Governmental appliances to help better policy discussions and country-wise implementation
  • 19.
    AIRSPY AIRSPY is adeep convolutional neural network framework, integrated with a front-end user-ergonomic mobile application, to detect and predict defined air quality parameters using satellite remote sensing image data We thoroughly believe that AIRSPY will be a game-changer in timely detection and prediction of air quality of our surrounding atmosphere with a fast, accurate and fully- automated approach
  • 20.
    Nuwan Bandara Team Leader ProjectAIRSPY Team Phantom Promod Fernando Team Member Kalanaka Samarasinghe Team Member Sahan Hettiarachchi Team Member
  • 21.
    THANK YOU Air Quality isthe crucial substance of our whole living stream and thus, it must be our collaborative effort to protect our atmosphere as far as our ability since, " A LIFE DOES NOT EXIST WITHOUT AIR"