Data Analytics

Landscape

6/25/22 11:59 AM 1
Smart City & Internet Of
Things

2
Agenda

1 Smart City

2 Internet of Things

3 Data

4 Role of Big Data

5 Role of Cloud Technology

Definition
A resilient, inclusive and collaboratively-built city that uses technology and data
to better the quality of life for all people.
What is Smart City?
• Enabling supply & demand size policy
• Transparency & Open data
• ICT & eGov
Governance

• Entrepreneurship & innovation

• Culturally vibrant & happy • Productivity
• Safe Living Economy • Local and global
• Healthy interconnectedness

Smart
•
City
21st century education • Mixed Model Access
• Inclusive society People Mobility • Clean & non motorized option
• Embrace Creativity • Integrated ICT

Environment
• Green Building
• Green Energy
• Green urban Planning
Definition of Smart City
• Reduction of CO2 emissions; Use of renewable energy
Environment sources, Green building, Green urban planning.

• Co-working, cultural vibrant, living-lab, crowdsourcing,

Living healthy, safe.

• Development of technologies to improve urban mobility, low

Mobility environmental impact, non-motorized options, autonomous.

• Enabling supply & demand size policy, Transparency & Open

Governance data, ICT & e-Gov.

• Entrepreneurship & innovation, Productivity, local and global

Economy footprint

• Sharing of data, security and protection of sources,

People networking and communication, 21st century education,
Inclusive society, Embrace Creativity
Smart City & Actor Map
Public
Universities / Health
Housing State/National
Schools City Care
Associations Government Energy
Government Providers Providers

Research
Institutions

Startup

User
Incubators Transport
Providers

Citizen
Digital
Agencies Investors
Banks/
Insurance
companies

Hotels,
Manufacturers & Logistics Technology Museums,
Retailers Construction Providers Vendors Telecom Restaurants,
Companies Providers Theaters,
Stadiums
Technologies & social innovations
Drones

Internet of Things 3D printing

Renewable Energy

Crowd sourcing

Social Robotics Block Chain

Gamification
Mobile

Sharing economy

Co creation
Social Media

Cloud
Big Data Self Organization

Artificial Intelligence
Smart City Challenges

Technological
challenges
Source of Cost of Cloud • Privacy
Business
Planning Sustainability markets and acquiring computing • Data Analytics
Challenges • Data Integration
customer smart city integration
• GIS-based
Visualization
• QoS
E-Governance

E-
Governance

Online Complaint Payment

Records Management Gateways
Smart Home

Smart
Smart Media
Air Control
Appliances

Lights & Background

Curtains Music

Central
Central AC
Heating

Video
Safeguard
Intercom
Smart Grid

Nuclear Power Plant Hydraulic Power

Generation

Thermal Power Plant

Photovoltaic Factories

Cities & Offices

Homes

Ecological
Vehicles

Wind Generator
Summary Of Different Smart City Applications
Possible
Application Specific Use Internet of Things Communication Advantages Limitation
Technologies
Smart Healthcare Health monitoring. Sensors, smart Bluetooth and • Early diagnose Lack of Precision
wearable devices, ZigBee the disease
Smart Transport Efficient route smart cars, cameras, RFID< 3G, and 4G • Automatic traffic Network
management RFID cards management disconnectivity can
• Efficient route cause serious
management accidents
• Less Congestion
Smart Governance To make smart Smartphones, Wifi, LTE-A, WiMax, • Awareness in Collection and
policies with the aim cameras, sensors Bluetooth, LoRaWan, terms of citizens analysis of data
of managing the needs seem difficult task
citizens
Smart Grid To manage the Smart meters and Wifi, Zigbee, Z-Wave • Clear policy • Costly
power supply Smart readers • Efficient power • Hard to manage
supply
• Future needs
estimation
Technology – Smart City

Technology Mission
Network For gathering data
Field Gateways For Facilitating data gathering and compression
Cloud gateway For ensuring secure data transmission
Streaming Data For aggregating numerous data streams and distributing them
Processor to a data lake
Data Lake For storing data the value of which is yet to be defined
Data Warehouse For storing cleaned and structured data
Data Analytics Tools for analyzing and visualizing data collected by sensors
Machine Learning For automating city services based on long-term data analysis
User Applications For connecting smart things and citizens
Smart City Case Studies

City Of City Of Helsinki City of

Stockholm • Publicly 1030 Copenhagen
• Smart Waste opened Databases • Innovative
Management related solutions related
• Smart Traffic Transportation, to transport,
Management Economics, waste
Employments etc. management,
water, heating etc.
Agenda

1 Smart City

2 Internet of Things

3 Data

4 Role of Big Data

5 Role of Cloud Technology

The Internet of Things

Objects with sensing or

processing capability, each
of which is assigned a unique
IP address, that
communicate data over the
internet without the need for
human interaction
Some Applications of Internet of Thins

Manufacturing Smart Building

Smart Cities Energy

Internet
of
Home Automation
Things Environment

Logistics Agriculture
IoT Platforms
IoT Platform Today IoT Platform in the future
Due to the different protocols and standards, devices are very much Developments on IoT platforms indicate a more comprehensive
specialized to capture information in specific characteristics creating a connectivity enhancing the data collection to enable true creation of
chain that is hermetic to certain applications although the use of APIs value. Some application might still require a separated platform but
can improve significantly its use. In Practice, IoT platforms today is that would be by design instead of lack of options.
segmented.

Analytics
Analytics

IoT Platforms
IoT Platforms

Private
Consolidated
platform by
Platform
design

Communicati
Communicati

on Protocol
on Protocol

Protocol
Protocol

Protocol
Protocol

Protocol

Specialized
Specialized

Gas , Gas ,

Sensors
Sensors

Smart Smart heating & Smart Smart Smart heating & Smart
Securing Securing
Electricity Traffic water cleaning Electricity Traffic water cleaning
supply supply
Too much focus is on The THINGS, not the ANALYTICS

INVESTMENT FOCUS VALUE FOCUS

Data Store

Sensors Edge Integrate

Analytics
System of Systems
ACTING THINKING
OPTIMIZING
Agenda

1 Smart City

2 Internet of Things

3 Data

4 Role of Big Data

5 Role of Cloud Technology

Data is the lifeblood of a smart
city
Three Layers of Data

First is the technology base, which includes a critical

mass of smartphones and sensors connected by high-
speed communication networks.

The second layer consists of specific applications.

Translating raw data into alerts, insight, and action
requires the right tools, and this is where technology
providers and app developers come in.

The third layer is usage by cities, companies, and the

public. Many applications succeed only if they are
widely adopted and manage to change behavior.
Open Data
Open data is machine readable digital data, freely available for everyone to use and republish.

In Smart city data is as important as the physical infrastructure. In the

past centuries, cities developed carefully thought our mechanisms to
manage the physical infrastructure is required to maintain data and
share data and hast the following features:
• Authority: the data infrastructure is credible, authoritative source
of data.
• Openness: The data infrastructure is open for all users, making
the data as accessible as possible.
• Real Time: The value of data decreases with its age. The use of IoT
generates massive volumes of real time data which can be made
available though the data infrastructure.
• Agility: In addition to the previous aspect, agility to update data
quickly, the data infrastructure should also be agile to encompass
new data sets.
Agenda

1 Smart City

2 Internet of Things

3 Data

4 Role of Big Data

5 Role of Cloud Technology

Role of Big Data In Smart Cities

Sensors/Devices
Internal Data
Sources

Open Data

External Data Capture > Compute > Analyze > Store Smart Solutions
Citizen Data
feeds
Key Requirements For Big Data Platform

Security and
privacy

Intelligent and
Scalable dynamic

Distributed and
Real-time decentralized
Landscape of Smart City & Big Data Technologies
Smart Industry Smart Grid

Smart Police

Smart Health
Query Engine e.g. Hive, Mahout
Smart People

Programming model for processing large

data sets with a parallel, distributed
algorithms

Distributed fault tolerant database for

Management
Smart Waste
large unstructured data sets like NOSQL.
Smart Home

Smart Parking

Smart Traffic
Construction frame of Big Data Technologies For Smart City
Example of some Application Services

Analytics/Reports Recommender Fraud Detection Sentiment Analysis Intelligent Traffic Management Web Display Advertising

Smart Analyzer

Decision Making Programming model for processing large data sets with a parallel, distributed algorithms

Split Map
Split Map Reduce Output
Input Shuffle Sort
Split Map Reduce Output
Split Map

Distributed fault tolerant database for large unstructured data sets like NOSQL.

Source of Big Data

Smart Environment
Business Model for Big Data & Smart City

Customer Decision Makers City Managers Practitioners

New Market Approach Alerts Reports Dashboard

Decision Data Driven Business Performance

Predictive Analysis

Data Science
Automation Business Management
Security

Business Intelligence Application Business Intelligence Application

Cloud
Smart Devices Unstructured Big Data Data Warehousing
Agenda

1 Smart City

2 Internet of Things

3 Data

4 Role of Big Data

5 Role of Cloud Technology

Convergence of IoT, Big Data and Cloud

For IoT, number For Big Data, For Big Data, Cloud offers IoT is a King, Big
of billions of data collection is data is nothing Everything as a data is a Queen
connected one of the main without real Service business and Cloud is a
devices is an concern, and IoT business value model for IOT Palace
indicator of IoT. can play an insight and big data.
The connectivity important roles
is just an enabler for data
but the real value collection and
of IoT is on data data sharing
(business
insight/data-
driven economy)
iKaaS-EU-Japan IoT big data platform
App. App.

Query

Storage KaaS
Global Cloud

Data Data
Knowledge
Knowledge
Security Security
GW GW

Query Query

Local Cloud Local Cloud

Storage Storage
Data Data

Sensors Sensors
/IoT Devices /IoT Devices

The goal of this project is to combine ubiquitous and heterogeneous sensing, smart objects, semantic, big data and
cloud computing technologies in a platform enabling the Internet of Things process consisting of continuous iterations
on data ingestion, data storage, analytics, knowledge generation and knowledge sharing phases, as foundation for cross-
border information service provision.
Cloud Based IoT Big Data Applications

Massive monitoring

Prescriptive and descriptive

Real-time actionable insight

Proactive and predictive

Smart City & Internet Of Things, Data, Services, People

Internet of Smart Smart Internet of

Things people Mobility Services

Smart Smart Smart

Living City Environment

Internet of
Internet of Smart Smart People
Data Governance Economy
Video Analytics Using
Deep Learning
Neural Networks in Our Daily Life

Mobile Search Self Driving Cars Computer Gaming

Processing
Navigations
Images in Phones
What is Neural Network?

Neural network is a system

or hardware that is
designed to operate like a
human brain.
What Can Neural Network Do?

Recognize
Speech
Read
Identify
Handwritten
Faces
Text

Translate Neural Control

Text Network Robotics

And Lot of Other

Things
Artificial Neural Network

W1 W2

• 3-layer neural network with one input

layer, one hidden layer, and one output
layer.
• The number of nodes in the input layer
is determined by the dimensionality of
our data
• The number of nodes in the output layer
is determined by the number of classes
we have Input Layer Hidden Layer Output Layer
How Does Neural Network Works?
Wonder how traffic cameras identify
vehicles on the road to detect
speeding vehicles and those breaking
the law…!!
How Does Neural Network Works?

28 * 28 Pixels of the image is

fed as input to identify the
registration plate

Let’s consider the image of this

vehicle and find out what's on the
number plate.
How Does Neural Network Works?

Each neuron has a number

called activation that
represents
Each Neurontheisgrayscale value
lit up when its
ofactivation
the corresponding
is close topixel
1
ranging from 0 to 1. 1 for white
pixel and 0 for black pixel
How Does Neural Network Works?
Pixels
The
The in the
weights form
are
interconnections
Lets Name of inputs
the arrays
multiplied
are are X2
with
assigned
X1, fedand
the toweights
input the
X3 input
singles
at
Finally,
The
As the
The The
weighted
input model
signal
layersum
flowswould
ofwithin
passes predict
inputs
it the
to isthe the
fed
hiddenasoutcome
hidden input
layers, by
to the
layers..the
and suitable
afunction layer..
biasrespectively..
isactivation
added
random to all of them..
applying
activation
weighted sum of inputsto decide function
which
is calculated to the
nodes
and to
is fed
to the activation output
featurelayer
fire forfunction extraction..
of each layer to decide
W2
W11 1
which nodes to fire.

W1 W2
2 2 Activation
Function
W1 W2
x1 3 3 W21

W1 W2 W22
x2 4 4
W23
W1 W2
5 5 W24
x3
W1 W2
6 6

Pixels of car W1 W2 Identifies the number

7
image fed as 7
on the plate
input W1 W2
8 8

Optical character Recognition (OCR) is used on the TheWith

output is compared
every with
iteration, the the original
weights result
at every
images to convert it into text in order to identify what's and multiple iterations
interconnections are done
are adjusted to get
based onthe
the
maximum accuracy.
error…
written on the plate
Activation Functions

Function Name When it used

Threshold Function Used when the output depends on a
threshold value
Sigmoid Function Used when the model its predicting
probability
RELU function It gives an output X if X is is Positive, o
otherwise
Hyperbolic tangent function Similar to sigmoid function a range of (-1, 1)
Back Propagation
Error in The Output is backpropagated
through the network and weights are
W2
adjusted to minimize the error rate. This is
W11 1 calculated by a cost function.
W1 W2
2 2 Activation
Function
W1 W2
x1 3 3 W21

W1 W2 W22
x2 4 4
W23
W1 W2
5 5 W24
x3
W1 W2
6 6

Pixels of car W1 W2 Identifies the number

7
image fed as 7
on the plate
input W1 W2
8 8

Backpropagation
Limitations of Machine Learning 1/2

Machine Learning

Are not useful while working

with high dimensional data, Cannot solve crucial AI
that is where we have large problems like NLP,
number of inputs and Image recognition etc.
outputs.
Limitations of Machine Learning 2/2

• One of the big challenges with traditional Machine Learning models is a process
called feature extraction.
• For complex problems such as object recognition or handwriting recognition,
this is a huge challenge
Deep Learning to The Rescue
• Deep learning models are capable to focus on the right features by themselves, requiring little
guidance from the programmer.
• These models also partially solve the dimensionality problem.
• Deep Learning is implemented through Neural Networks.
• Motivation behind Neural Networks is the biological Neuron.

What is
Deep
Learning?

The idea behind Deep Learning is to build

learning algorithms that mimic brain.
What is Deep Learning

More accurate
Type of Machine Requires less data results than
Learning preprocessing traditional machine
learning

Can process huge Suitable for images

Deep Learning
amount of data and processing, video
network consists of
find complex analytics, voice
multiple layers
patterns in it. recognition etc.

05/18/2025 07:38 AM 51
What is Deep Learning?

A collection of statistical machine learning techniques used to learn

feature hierarchies often based on artificial neural networks.

Input Layer

Hidden Layer
1

Hidden Layer
2

Output Layer

Input Layer Hidden Layers Input Layers

Picks up input signals Does All type of calculations & This layer delivers the
and passes them to Feature Extraction final result
the next layer
Deep Learning Example
Patterns of
Local Contrast
Face
Features

Face

Output Layer

Hidden Layer 2

Hidden Layer 1
Input Layer
Deep Learning Applications

Self Driving Cars Voice Control

Assistance

Auto Image Caption Automatic Machine

Generation Translation
Deep Learning Can often Outperform Traditional Methods
Deep Learning can often outperform traditional methods % reduction in error rate achieved
by deep learning vs traditional methods

Image Classification 41%

Facial recognition 27%

Voice Recognition 25%

0% 5% 10% 15% 20% 25% 30% 35% 40% 45%

05/18/2025 07:38 AM 55
Object Detection Use
Case
Use Case Statement

Aim of the use case is to build a

Video-Analytics Dashboard which
gives information for different
objects like Cars, Trucks,
Pedestrians detected object from
different video streaming sources.
Use Case Overview

This Use-Case aims to use the existing infrastructure by taking the raw
video streams from camera translating them into real data on which
we can make smart decisions.
Following are some of the Use-Cases for Traffic management using
Video-Analytic
• Vehicle and Pedestrian detection.
• Congestion Detection.
• Vehicle count.
• Footfall analysis.
• Vehicle classification.
• Anomaly detection.
Video Analytics Pipeline
Generated Images Image
Annotation
Using Labeling La
b
Video Sample int ele
o t d Im
he a
Tr Mo ges
ain de F
i ng l F e d
or

Cluster with

Training model GPU

GPU

Enabled Cluster
+

TensorFlow Serving

Dashboard showing different Expose the Mode results as

statistics and analytics results REST-API

REST API
Video Analytics Pipeline

• Model-Training: A pre-trained deep-learning model has been used. The model is

trained by ingesting the labeled images obtained from the video from CCTV and
other video streaming sources.

• Model-Serving: The model results will be served as a rest-API using Flask.

• Video-Analytics Dashboard: The model results from the Flask-API are stored
inside MongoDB/PostgreSQL, and served onto the Dashboard.
Technologies Used

• Tensorflow (https://www.tensorflow.org/ )

• Open-CV (https://opencv.org/ )

• Flask (http://flask.pocoo.org/ )

• MongoDB (https://www.mongodb.com/ )
What is OpenCV?

• Open source Computer Vision Library

• Originally developed by Intel
• Has more than 2500 optimized algorithms
• C/C++/Python API
• Cross Platform
• Also available for android and IOS
• Released under a BSD license (its free)
Applications of OpenCV

Human-Computer Object Object Face Recognition

Interaction (HCI) Identification Recognition

Gesture Motion Tracking Image Processing Mobile Robotics

Recognition
What is Flask?

• Flask is light weight python Micro Framework.

• A web framework where you can build API,
sites or web application is a very “Explicit Way”.
• Easy for beginners
• Reusing Existing code
• Neat way to write small applications
Image Labeling Tools

RectLabel Annotorious
• An image annotation tool to label images • Annotorious is an Open Source image
for bounding box object detection and annotation toolkit written in javascript;
segmentation. • https://annotorious.github.io/
• https://rectlabel.com/

LabelMe Yolo_Mark
• It provides an online annotation tool to • GUI for marking bounded boxes of objects
build image databases for computer vision in images for training neural network Yolo
research v3 and v2
• http://labelme.csail.mit.edu/Release3.0/ • https://github.com/AlexeyAB/Yolo_mark

Fast Image Data Annotation Tool

• A tool using OpenCV to annotate images
for classification, OCR, …
• http://tinyurl.com/y28rpqfz
What is Labelimg?

LabelImg is
graphical image
annotation tool.

Qt (pronounced "cute“)is a free and open-source widget

It is written in toolkit for creating graphical user interfaces as well as
Python and cross-platform applications that run on various software and
uses Qt for its hardware platforms such as Linux, Windows, macOS,
graphical Android or embedded systems with little or no change in the
interface.
underlying codebase while still being a native application
with native capabilities and speed.

Annotations are •
The PASCAL file format type is primarily associated with
saved as XML Apple II operating system.
files in PASCAL • VOC files can contain audio clips, music files, instrument
VOC format,
sounds and sound effects that are used with the
Labelimg the format used
by ImageNet. associated hardware device. The VOC file format is a
compressed file format, allowing for a reduction in the
size of the audio file that is stored on the user's hard
It also supports drive.
YOLO format. You only look once (YOLO) is a state-of-the-art, real-time object detection
system. On a Pascal Titan X it processes images at 30 FPS and has a
05/18/2025 07:38 AM
mAP of 57.9% on COCO test-dev.
66
What is NVIDIA’s GPU?

GeForce RTX 2080 Ti. NVIDIA's newest

flagship graphics card is a revolution in
gaming realism and performance.

NVIDIA GPU has advance architecture,

breakthrough technologies, and 11 GB of
next-gen, ultra-fast GDDR6 memory

It is the world's ultimate gaming GPU.

Nvidia GPU

Developers & machine learning enthusiast

loves this GPU
Why GPU Is Used For Deep Learning?

• In above figure there is comparison of throughput of different neural networks

with processing capability.
Image Labeling Tools

CPU GPU
Really fast caches (great for data Lots of math units
reuse)
Fine branching granularity Fast access to onboard memory
Lots of different processes/threads Run a program on each
fragment/vertex
High performance on a single thread High throughput on parallel tasks
of execution

CPUs are great for task parallelism GPUs are great for data parallelism
CPU optimized for high performance GPU optimized for higher arithmetic
on sequential codes (caches and intensity for parallel nature (Floating
branch prediction) point operations)
Deep Learning & machine learning Performance Comparison

Deep learning outperformed as compared to neural networks & traditional

machine learning.
Why Only Deep Learning is Used For Video Analytics?

• Computer vision community always struggle with

finding features that can work in different
conditions, such as illumination, camera angles,
background, crowded scenes etc.
• Deep Learning provided an opportunity to learn
generic features, rather specific ones.
• Combining other data augmentation tricks such as
cropping, flipping, shearing, zooming, changing
lighting conditions etc, one can learn good
hierarchical representations using Deep Learning
that may be helpful in their discrimination (in case
of supervised learning).
QUESTIONS & ANSWER SESSION

05/18/2025 72