1 of 1

[1] G. D. Jyothi and K. Navya,” Design and implementation of a store management system,” 2017 International Conference on Intelligent Sustainable Systems (ICISS), 2017, pp. 1149- 1151, doi: 10.1109/ISS1.2017.8389365.

[2] A. Marin-Hernandez, G. de Jesus Hoyos-Rivera, M. Garc´ ´ıa-Arroyo and L. F. Marin-Urias, ”Conception and Implementation of a Supermarket Shopping Assistant System,” 2012 11th Mexican International Conference on Artificial Intelligence, 2012, pp. 26-31, doi: 10.1109/MICAI.2012.21.

[3] M. Serasinghe and S. Vasanthapriyan, ”Intelligent Retail Checkout Management System,” 2016 2nd IEEE International Conference on Computer and Communications (ICCC), 2016, pp. 929-932, doi: 10.1109/CompComm.2016.7924840.

[4] M.Kemp, “People waiting in line with shopping baskets at grocery store”, Accessed May 11, 2023 Available: https://www.gettyimages.com/detail/photo/people-waiting-in-line-with-shopping-baskets-at-royalty-free-image/74214360?adppopup=true

[5]Jes, “Supermarket aisle with empty red shopping cart background” Accessed May 11, 2023, Avaialble:https://depositphotos.com/123614652/stock-photo-supermarket-aisle-with-empty-red.html

Autonomous Customer Assistance Tool

Project Advisor: Prof Andrew Bond

Introduction

E-commerce is popular but physical stores still attract customers. Why?
In-store shopping allows customers to try and test products before buying, avoiding shipping costs and wait times.
Consumers today are well informed and have higher expectations; expect personalized experiences, and good customer service.
Major issues in physical stores: Trouble finding product prices, product availability, long lines and wait times.
For the stores to provide seamless customer experiences, improving management systems is must.

Methodology

Analysis and Results

Summary/Conclusions

Key References

Acknowledgements

This web application is a powerful tool that can help retail managers optimize their staff allocation and improve the customer experience.
Machine learning techniques to analyze customer density and behavior, the application can provide real-time insights that allow managers to allocate staff more efficiently and effectively, reducing wait times and improving customer satisfaction.

Apart from the efforts put in by us, the success of this project depends largely on the encouragement and guidelines of many others. We would like to pay gratitude to our coordinators, Prof. Dan Harkey, Prof. Andrew Bond for their constant encouragement and guidance.

Detecting Aisle Density

To allocate staff to an aisle, first step is to detect the number of people in that aisle :

Read Frame

Used CV2 python libraries to read frames from video camera and images.

Preprocessing

Frame is resized. Then converted to gray scale to eliminate the complexities. Lastly, image scale and distance between pixels is set for higher accuracy.

HOG Descriptor & SVM

Feature descriptor to detect people in an image using Support Vector Machines.

Design

Architecture

Auto Assist adopts service-oriented architecture for independent deployability, and reusability. RESTful APIs are used to make asynchronous https calls to different services.

Version control: Using Git to manage changes to the application code, infrastructure configuration, and deployment scripts.
Cloud Platform: Using AWS to deploy our service oriented application, enabling different components/services to interact with each other.
Monitoring: Using AWS to monitor real-time logs, metrics, and event data of our application.

Computer Engineering Department

Gupta, Akanksha (MS Computer Engineering) Joseph, Ashly (MS Computer Engineering)

Patel, Bhaummi (MS Computer Engineering)

Shukla , Raksha (MS Computer Engineering)

Current solutions like call buttons are not effective.
Our solution improves customer service by providing real-time assistance using AI to make informed decision in staff allocation.

Component Design

1. Notifications

In response to receiving the aisle density, the store manager assigns an employee to a particular aisle. When an employee is assigned an aisle he/she will get notified. This is implemented using WebSocket.

2. Messages

Messages is a simple, helpful messaging features that keeps the store staff including the manager connected with the other staff. This is implemented using WebSocket.It allows the staff to communicate in real-time which helps in responding quickly to queries and addressing any potential issues swiftly. Messages also ensures that everyone stays informed and up-to-date.

Technologies Used

Cascade Classifiers		HOGDescriptor
fullbody_detector	profileface_detector	HOGDescriptor
55%	62.5%	87.5%

Using a machine learning model with good performance is crucial to the success of the application.

Precision�Testing various human detection algorithms proved HOGDescriptor to be the most effective for Auto Assist. The precision score of top performing Machine learning algorithms for our application.

Deployment and Operations

Below are the images (obtained from getty images [4],[5]), that show the performance of HOGDescriptor, the green rectangles implies the positive identification of a person.

PeopleDetected = 0

PeopleDetected = 5