REAL-TIME CROWD DETECTION ANALYTICS IN TRADING OUTLETS

7 min readDec 24, 2021

INTRODUCTION

The crowd is a significant source of the transmission of the COVID-19 spread. Prevention and alleviation measures have focused on reducing people’s mass gatherings. When it comes to the Covid-19 pandemic, communities are still unsure how to resume a “new normal” existence while the virus is still circulating among the people, despite an initial lock-down phase. The hope of universal vaccination represents a massive challenge for humanity, and it is not likely to be achieved even within months. Meanwhile, the need of the hour is to develop a mechanism that will function towards bringing normalcy in the lives of people while limiting the spread of the virus. Trading outlets, like shopping malls, restaurants, etc., caters to the mass gathering, which may result in the exponential spread of the virus. To avoid such risk, we may use the concept of “Counting” and let people know the real-time Crowd Count of these outlets. Crowd Count will help you know how many people are present in the outlets leading to effective monitoring and management of crowd levels.

NEED AND MOTIVATION

Need

The need to implement the project is to prevent the spread of coronavirus in trading outlets. Outlets can use this technology to detect crowd count and provide this information to the public or customers. The project focuses on developing a web application that provides a live crowd count that will help the customer to get to know the real-time count of the people which will allow the customers to make a rational choice as to whether to visit the outlet at the moment or schedule it for a later time period as to avoid overcrowding. This will also allow managing their time in an efficient manner. Crowd Count will also help outlets to manage and regulate occupancy and detect overcrowding. The most significant and peculiar precaution this global pandemic has given rise to is the fact that each and every individual needs to ensure that they maintain advised social distancing among themselves and crowd detection can ensure that social distancing follows. Crowd Counting technology can be implemented by using cameras installed at entrances and exits of the outlet to assist in the enforcement of live crowd count and Web Technology can be used to make the crowd count data available to the customers in real-time. While this method does not directly address the physical separation of individuals, it is a step in the right direction.

Motivation

Motivation for doing this project came from a real problem that is not solved yet. E.g., we have various shopping centers like D Mart, Big Bazaar, etc. Pre-covid people used to visit them in large numbers. Eventually, these trading outlets became the first preference for customers to buy groceries or any other necessary items for their use. Now during the covid era, these outlets, the primary choice of the masses, people still want to visit them but are scared of covid and mass gathering. However even if the individual decides to visit such mega stores, upon reaching the store, they are forced to abandon their visit due to the fact that the place is overcrowded and lacks social distancing. This problem motivated us to do a project that can count the crowd in real-time at these outlets and display the information on the respective outlet’s website.

PROBLEMS AND OBJECTIVE

Scope and Beneficiary

The project intends to cover the general public and outlet companies. We will be utilizing the cameras that are pre-installed in the outlets to detect the person, count the number of persons that are currently in the outlets, and, using the outlet’s website we will display the information to the general public.

Using this project, we are able to benefit both public and outlet companies. By examining the real-time crowd count at the outlet’s website, people will be able to make a decision as to when to visit the outlet and avoid it in case of overcrowding, leading to less risk of covid in the time of the global pandemic. Outlet companies will also be benefitted as they will get the actual number of people and will be able to handle the crowd accordingly as per the social distancing norms. Further, they can also use the data for customer behavior analysis.

Objectives

1. Counting The Number Of People Currently Present In The Outlet

The primary goal of this project is to provide machine learning functionality to pre-installed CCTV cameras at outlets to detect people count using Computer Vision Technology (Object Detection).

2. Real-Time Crowd Count Statistic On Website

The project also aims to provide crowd count on the website, which will help the end customers know the number of people possibly in the outlet in real-time.

3. Gender And Age Estimation

Estimation of age and gender of customers getting in the outlet will help the outlet companies to examine the type of crowd that comes in their outlet and provide services accordingly.

4. Helping Outlets In Handling Crowd Management

Real-time crowd count will help the outlets to design strategies for the future and use this data to effectively predict and forecast the crowd count. Outlets will also be able to handle the crowd in case of an overcrowding scenario. Further analysis like what type of gentry is entering in the outlet, what is the peak hour etc. can also be taken into consideration for building future strategies.

5. Helping the public to make an informed decision

The project is helping to get to know the real-time count of the people inside the outlet, which will lead to the customer’s choice whether to visit it now or schedule it later in case of overcrowding. This will save the customer’s time considering the fact that time is precious. Because of this, the public will be able to make an informed decision.

TOOLS AND TECHNOLOGIES

The major technologies used in the project is:-

a. Computer Vision (Object Detection)

b. Web Technology

Computer Vision (Object Detection)

Python libraries and Algorithms for Object Detection, Object Tracking (people), Age & Gender estimation are

1. Numpy

2. OpenCV

3. Dlib

4. imutils

5. Pre-trained Caffe Deep learning models

The algorithm used for Object Detection(people) is a hybrid approach MobileNet-SSDs+Linear SVM as it is a highly accurate object detection method without as much of the computational burden. For object tracking, we have used a ‘centroid object tracking algorithm’ in order to track the object in a video stream.

Web Technology

The project will be using the website as the delivery mode. Crowd Count, Age & Gender of people will be stored in a database, and later, using the website, we will display the real-time information to the public.

1. Front-end Technology (HTML, CSS, Bootstrap)

2. Back-end Technology (Django)

3. Database (MongoDB)