SCHOOL OF ADVANCED SCIENCES

Project - Introduction to Innovative Projects (PHY 1999)

It is certified that the project entitled “Smart Disaster Management” is the Bona fide

work for Project component of Introduction to Innovative Projects by the following students

Amartya Gaur(17BCE0414)

Nitisha Singhal(17BCE0050)

Shreyansh jain(17BCE0292)

Rahul Kumar(17BCE0222)

Lakhan Solanki(17BCE0309)

Pursharth Nohria(17BCE036)

Tanvi Gupta(17BCE0909)

Priyanshi Chauhan(17BIT0071) Anish Abhijit Diwan (17BME0109) of CSE/IT/MECH

branch under my supervision in TG1 slot during the Winter Semester -2017 at V.I.T. University,

Vellore-632 014.

Faculty Signature:

Date : 01/04/2019
1. Introduction
A crowd is a large group of people that are gathered or considered together. This bunch of people

can be gathered anywhere like common tourist places, pilgrimage destinations etc. Large

gathered people are often difficult to manage especially in the case of disaster situations. A

disaster can be a sudden, calamitous event that seriously disrupts the functioning of a community

or society and causes human, material, and economic or environmental losses that exceed the

community’s or society’s ability to cope using its own resources. These disasters can be natural

or human based. Proper management is required at the time of the disaster. Disaster Management

can be defined as the organization and management of resources and responsibilities for dealing

with all humanitarian aspects of emergencies, in particular preparedness, response and recovery

in order to lessen the impact of disasters. There are various ways of disaster management. Any

problem can never be solved completely but its effect can be reduced. If we in prior know the

amount of people present in the mass gatherings disaster can be managed efficiently .In case of

disaster special relief camps can be arranged and proper help can be provided including food and

everything .The smart disaster management technique helps to tackle this really well. The plan of

our project is to build a mobile application that can estimate the number of people in the crowd

through image recognition technique and then the concerned authorities can deal accordingly

with the number.

2. Objective of the project

to people being trapped in large masses. A disaster in a heavily populated region

would mean that relief forces will have to dispatch an unknown amount of food,

medication, clothes and other relief items.

This havoc can be immensely reduced if the number of people in the crowd can be

estimated.

The project aims at designing a neural network which can estimate the number of

people using an image of the crowd.

We can take a picture of the crowd and count the number of people and can provide

the relief to them in case of the disaster.

3. Definition of the Problem

There are multiple mass gatherings in the world. These can be exclusive international fairs,

Pilgrimage destinations or daily destinations including airport, railway stations, malls etc.

When these gatherings are coupled with an unfortunate event of occurrence of the disaster (it

may be a stampede due to the fire alert or it may be natural disaster occurring in pilgrimage

destinations including floods and earthquake, or some other disaster), it leads to havoc, it

sometimes leads to death. So we are developing a solution to deal with disaster involving

smart disaster management techniques of Crowd counting. For example kumbh ka mela,is a

fair held every 12 years in Prayagraj in India is a huge gathering. It attracts lacs of people

from different corners of country and the world. The safety of those people comes in the

hands of government,police force and disaster management team. It is not easy for them to

keep the count of people in such a gathering. So our project has proposed and developed the
 idea of smart disaster management using crowd counting approach which helps the disaster

management teams and other relief forces to count the crowd easily and in proper way so that

in case of disaster they can provide relief to the affected people easily.

4. Methodology & Experiment

The application would contain the details of all mass gatherings going to happen in the year

especially in the accident prone areas. At the time of the event images would be gathered

from various sources including media and satellite images. Each image would be used to

predict the number of people gathered and then the average of all shall be taken to estimate

the total number of people gathered. That would be sent to the concerned disaster

management to take appropriate measures for preparing for such an unfortunate event. The

application developed would also have features like distress call which would allow a person

stuck to alert the rescue team of his/her location.

Algorithms used:

Convolutional Neural Network(CNN)

A Convolutional Neural Network is a Deep Learning algorithm which can take in an input

image, assign importance to various aspects/objects in the image and be able to differentiate

one from the other. The pre-processing required in a ConvNet is much lower as compared to

other classification algorithms. While in primitive methods filters are hand-engineered, with

enough training, ConvNets have the ability to learn these characteristics.

The architecture of a ConvNet is analogous to that of the connectivity pattern of Neurons in the

Human Brain and was inspired by the organization of the Visual Cortex. Individual neurons
respond to stimuli only in a restricted region of the visual field known as the Receptive Field. A
collection of such fields overlap to cover the entire visual area.
 Markov random fields

Bayesian networks are a class of models that can compactly represent many interesting

probability distributions. However, some distributions may have independence assumptions that

cannot be perfectly represented by the structure of a Bayesian network. In such cases, unless we

want to introduce false independencies among the variables of our model, we must fall back to a

less compact representation (which can be viewed as a graph with additional, unnecessary

edges). This leads to extra, unnecessary parameters in the model, and makes it more difficult to

learn these parameters and to make predictions. There exists, however, another technique for

compactly representing and visualizing a probability distribution that is based on the language of

undirected graphs. This class of models (known as Markov Random Fields or MRFs) can

compactly represent independence assumptions that directed models cannot represent.

Resnet

A deep residual network is a type of specialized neural network that helps to handle more

sophisticated deep learning tasks and models. It has received quite a bit of attention at recent

IT conventions, and is being considered for helping with the training of deep networks.

Datasets used:

UCF: This data set contains videos of crowds and other high density moving objects.

This data set contains images of extremely dense crowds.

ShanghaiTech Dataset

Technology stack:
 For Iterative Crowd Counting: Machine learning (neural nets) and image processing

wherein the crowd images will be obtained from sources like media, satellite.

For image classification: ResNet trained on the ImageNet dataset .

For Smooth Estimation count: Markov Random Field will be used to account for the

possible errors made while determining the counts for the patches.

For app and web development: a) Django as web backend, HTML, CSS for front end

b) REST framework for providing API’s c) JAVA and Swift for app development.

Snapshot of the working Model:

Working of app:
5. Summary of the results :

The demonstration of the working model clearly shows the number of people after entering the

image of the crowd.

6. Project conclusion:

The disaster can’t be predefined or can be stopped. So as they say ‘precaution is better than

cure’. So through our project the crowd can be counted and the relief can be provided on time

for those in need. Disaster Management can be summarized as the knowledge developed by

governments, professional response and recovery organizations/agencies, communities and

individuals to effectively respond to, and recover from, the impacts of current hazard

conditions. There are several ways by which disaster management can be achieved as discussed

above.These strategies help us to provide disaster relief i.e “The provision of emergency

services and public assistance during or immediately after a disaster in order to save lives,
 reduce health impacts, ensure public safety and meet the basic subsistence needs of the people

affected".We will be developing an app based on the mechanism on crowd counting.The target

audience can be places with mass gatherings,be it like fairs(kumbh mela),malls,railway stations

etc.The technique of crowd counting is used to count or estimate the number of people in a

crowd.By giving the number of people,the app helps the recovery organisations in the

management of the number of safety equipments required(food, life jackets,other help) which is

in turn contributing in smart disaster management.

Advantages:

As we mentioned before the kumbh ka mela. We all know that how big is that so through image

recognition we can count the number of people gathering there so that in case of disaster we can

inform authorities.

Crowd counting is one of the first and foremost parts of crowd management. It has several

realworld applications like crowd management, safety control and urban planning, monitoring

crowds for surveillance, modelling crowds for animation and crowd simulation.

The model will count the people in a building where the huge gathering is there. Eg like office

buildings where a lot of people work . so counting head will help in informing the authorities if

any disaster take place.

Disadvantages:

The counting cannot be perfectly accurate because it will be totally machine work.
 Improvements :

The above drawback or disadvantage can be covered by using proper quality hardware

where the image quality is really high.

7. Reference:

https://ermongroup.github.io/cs228-notes/representation/undirected/

https://en.wikipedia.org/wiki/Markov_random_field

http://www.cs.cmu.edu/afs/cs/user/bhiksha/WWW/courses/deeplearning/Fall.2016/pdfs/L

awrence_et_al.pdf https://github.com/desenzhou/ShanghaiTechDataset

https://www.crcv.ucf.edu/data/UCF101.php

8. Acknowledgement:

We would like to thank our professor for his continual help and support which led us to complete

the project on time. We would also like to thank our friends for helping us learn new skills to

make our project better.