Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

International Journal of All Research Education and Scientific Methods (IJARESM), ISSN: 2455-6211

Volume 9, Issue 6, June -2021, Impact Factor: 7.429, Available online at: www.ijaresm.com

Building a Smart Cart System for Retail Stores using


IOT and Machine Learning
Geetansh Kalra1, Dr. Santosh Sonavane2
1
B. Tech Student, Symbiosis Skills and Professional University, Kiwale, Pune
2
Director, School of Mechatronics, Symbiosis Skills and Professional University, Kiwale, Pune

-------------------------------------------------------------------*****************--------------------------------------------------------------

ABSTRACT

The “Smart Cart System”, is a new consumer cart that is designed to help consumers for better shopping
experience. This system will help consumers to avoid standing in long lines and save a lot of time for payment of
their bills. Micro-controller and Camera will be installed in the cart which will take the help of Object Detection
{Machine Learning technique} in real time to automatically scan the product being entered into the cart. Also,
an LCD is used to display the product being placed in the cart and its price. Using the concept of IOT and AI it
can also recommend the consumer the products which he/she can buy based on the products which they have
kept in the cart.

The Recommendations can be shown on LCD too. Once the Consumer is done with the shopping, the bill will be
displayed on the LCD itself and they would be able to pay through cards or UPI methods on the cart itself. This
would also help us to support Digital India initiative. The concept will revolutionize the shopping experience of
the consumers and can also be used to improve the inventory management in retail stores. Introduction of this
smart carts in India can change the way Retail Sectors used to work

Keywords: IoT (Internet of things), Machine Learning, Object Detection, Recommendation Systems

INTRODUCTION

The machine-driven cart, “Smart Cart”, is an associated innovative client buying product that's designed to assist
shoppers in fast-track their looking expertise. The construct of this good cart can revolutionize the buying expertise of
each client. The construct won't solely revolutionize the looking expertise of the client however it'll even have a nice
impact on the search keepers. They're going to get real-time knowledge regarding the product being purchased and can
understand the stock of the product at every moment.

The construct can improve inventory management and client relationship management. The introduction of these good
carts is going to be quite appealing within the retail sector and can result in a replacement era of shopping for and
merchandising products. Since the most issue it covers is saving time for the client, it may be thought of as an honest
possibility for future retails.

Existing Smart Carts


Amazon Dash Cart: Amazon is increasing its real-world footprint [1] with another unconventional physical product:
a handcart. Whereas it actually feels like an associate degree aesthetic upgrade to your commonplace grocery cart, the
Dash Cart, as it’s referred to as is in truth a sensible version of the tried-and-true food transport vehicle. It’s equipped
with a touchscreen and different varied hardware parts to mechanically find what things you’re inserting within and
even what number of these things you’ve picked off the shelf. Once you’re done searching, you’re allowed to require
the cart through a special lane that checks you out digitally while not requiring somebody's cashier to ring you up

Superhii Smart Cart:Since establishment in 2015, Superhii Technology has been focusing on the research and
development of smart shopping cart. Through smart shopping carts, it creates the ultimate shopping experience for
consumers and collects all shopping behavior data of offline retailing. As early as 2017, Superhii Technology put the
smart shopping cart in Wu Mart.

After that, it has successfully established cooperation with Baifule supermarket, Haolinju supermarket, Hefei Baida
and other brands of supermarkets, and Superhii smart shopping carts appeared in more and more supermarkets. [2]

IJARESM Publication, India >>>> www.ijaresm.com Page 2578


International Journal of All Research Education and Scientific Methods (IJARESM), ISSN: 2455-6211
Volume 9, Issue 6, June -2021, Impact Factor: 7.429, Available online at: www.ijaresm.com

Comparison of existing Smart Carts

Table 2: Comparing Smart Carts

Features Amazon Cart Superhii Cart

Automatic Billing Yes Yes

Automatic Product Recognition Yes Yes

Product Recommendation Yes Yes

Help to find Product in store No Yes

Voice Controlled No No

HARDWARE SELECTION

Arduino UNO- ATmega328 Microcontroller:


The Arduino UNO [3] may be a microcontroller board supported by the ATmega328. It's fourteen digital input/output
pins (of that vi may be used as PWM outputs), vi analog inputs, sixteen megacycles per second quartz oscillator, a USB
association, an influence jack, ICSP header, and a push.

Fig. 1: Arduino UNO- ATmega328 Microcontroller

It contains everything required to support the microcontroller; merely connect it to a PC with a USB cable or power it
with an AC-to-DC adapter or battery to induce started. The UNO differs from all preceding boards in this it doesn't use
the FTDI USB-to-serial driver chip. Instead, it options the Atmega8U2 programmed as a USB-to-serial convertor.

Table 2: Arduino UNO- ATmega328 Microcontroller Specifications

Operating Voltage 5V
Input Voltage (recommended) 7-12V
Input Voltage (limits) 6-20V
Digital I/O Pins 14 (of which 6 provide PWM output)
Analog Input Pins 6
DC Current per I/O Pin 40 mA
DC Current for 3.3V Pin 50 mA
Flash Memory 32 KB of which 0.5 KB used by bootloader

IJARESM Publication, India >>>> www.ijaresm.com Page 2579


International Journal of All Research Education and Scientific Methods (IJARESM), ISSN: 2455-6211
Volume 9, Issue 6, June -2021, Impact Factor: 7.429, Available online at: www.ijaresm.com

SRAM 2 KB
EEPROM 1 KB
Clock Speed 16 MHz

Force Sensitive Resistor:


FSR's sensors permit you to discover physical pressure, squeezing, and weight [4].

Fig. 2. Fore Sensitive Resistor with a coin

The FSR is formed of two layers separated by a spacer. The additional one presses; the addition of these Active
component dots bit the semiconductor which makes the resistance go down.

Fig. 3. Fore Sensitive Resistor

FSR's are primarily a resistance that changes its resistive worth (in ohms Ω) counting on what proportion it's
ironed. These sensors' area unit fairly low price, and straightforward to use however they are seldom correct.
They additionally vary some from the detector to detector maybe 10%. thus, primarily, once you use FSR's you
must solely expect to induce ranges of response. Whereas FSR's are a dangerous selection for sleuthing precisely
what percentage pounds of weight are on them.

Table 3: Force Sensitive Resistor Specifications

Size 1/2" (12.5mm) diameter active area by 0.02" thick (Interlink


does have some that are as large as 1.5"x1.5")

Resistance range Infinite/open circuit (no pressure), 100KΩ (light pressure) to


200Ω (max. pressure)

Force range 0 to 20 lb. (0 to 100 Newtons) applied evenly over the 0.125
sq in surface area

Power supply Any! Uses less than 1mA of current (depends on any
pullup/down resistors used and supply voltage)

Standard LCD 16 x 2 Display:


A Liquid Crystal Display [5] screen is an associate degree electronic display module and encompasses a wide selection
of applications. A 16x2 liquid crystal display is an incredibly basic module and is incredibly unremarkable employed in
varied devices and circuits.

IJARESM Publication, India >>>> www.ijaresm.com Page 2580


International Journal of All Research Education and Scientific Methods (IJARESM), ISSN: 2455-6211
Volume 9, Issue 6, June -2021, Impact Factor: 7.429, Available online at: www.ijaresm.com

Fig. 4. Standard 16x2 LCD Display

A sixteen x two liquid crystal means that it will display 16 characters per line and there are two such lines. During this
alphanumeric display, every character is displayed in a 5x7 constituent matrix. The sixteen x two intelligent
alphanumeric matrix show is capable of displaying 224 completely different characters and symbols.
 Operating Voltage is 4.7V to 5.3V
 Current consumption is 1mA without backlight
 Alphanumeric LCD display module, meaning can display alphabets and numbers
 Consists of two rows and each row can print 16 characters.
 Each character is built by a 5×8-pixel box
 Can work on both 8-bit and 4-bit mode
 It can also display any custom generated characters
 Available in Green and Blue Backlight

SOFTWARE AND ALGORITHM


Arduino IDE:
The Arduino Integrated Development Environment (IDE) is a cross-platform application (for Windows, macOS, Linux)
that is written in functions from C and C++. It is used to write and upload programs to Arduino compatible boards, but
also, with the help of third-party cores, other vendor development boards.

Fig. 5: An image of Arduino IDE

LabelImg:
Is a free, open-source tool for graphically labeling images. It’s very easy to use and could be used to label hundreds of
images. A bounding box is made on the image and a label is provided to that image. We have used LabelImg to label
our food products which we would be using for our real-time object detection. A whole dataset was made using
labeling.

Fig. 6. Image on the left is without labeling, Image on the right is with labelling

IJARESM Publication, India >>>> www.ijaresm.com Page 2581


International Journal of All Research Education and Scientific Methods (IJARESM), ISSN: 2455-6211
Volume 9, Issue 6, June -2021, Impact Factor: 7.429, Available online at: www.ijaresm.com

Dataset
Table 3: Dataset Split

Set Images Percentage Images

Training Set 408 88%

Validation Set 39 8%

Testing Set 19 4%

YOLO Algorithm:
To understand the YOLO algorithmic rule [7] , it's necessary to determine what's really being expected. Ultimately, we
tend to aim to predict a category of associate objects and therefore the bounding box specifying object location. every
bounding box is represented exploitation four descriptors:

1: Center of the bounding box


2: Width
3: Height
4: Value cis corresponding to a class of an object

In addition, we've to predict the pc worth, that is that the chance that there's an associate object within the bounding box

Fig. 7. Bounding Box showing the probability of an object in an image

As we tend to mentioned higher than, once operating with the YOLO algorithmic rule we tend to don't seem to be
finding out fascinating regions in our image that would probably contain the associate objects. Instead, we tend to
rending our image into cells, generally employing a 19×19 grid. every cell is to blame for predicting five bounding
boxes (in case there's over one object during this cell). Therefore, we tend to reach an oversized variety of 1805
bounding boxes for one image.

Fig. 8. Image being passed through a Deep CNN and encoded after

IJARESM Publication, India >>>> www.ijaresm.com Page 2582


International Journal of All Research Education and Scientific Methods (IJARESM), ISSN: 2455-6211
Volume 9, Issue 6, June -2021, Impact Factor: 7.429, Available online at: www.ijaresm.com

Most of those cells associated with bounding boxes won't contain an object. Therefore, we tend to predict the worth of
pc, which serves to get rid of boxes with low object chance and bounding boxes with the very best shared space during
a method referred to as non-max suppression.

Fig. 9: Applying the Non-max Suppression and finding the perfect box which identify the object in given image

RESULTS

We have used Yolo V4 to identify objects in real time. The accuracy of Yolo V4 is more than 98% on the dataset and it
takes less than 1 second to identify the object.

Fig.10. Initial Stage, Cart welcomes the customer

Fig.11. We take an object and place it in the cart Fig.12. Cart identifies the object and
display its price.

All the items and their prices are being stored in the database.

Fig.13. Once done with the shopping the total is displayed.

After the customer is done with the shopping a QR code is displayed on the LCD. From which the customer can pay
through UPI methods on the Cart itself.

IJARESM Publication, India >>>> www.ijaresm.com Page 2583


International Journal of All Research Education and Scientific Methods (IJARESM), ISSN: 2455-6211
Volume 9, Issue 6, June -2021, Impact Factor: 7.429, Available online at: www.ijaresm.com

CONCLUSION AND DISCUSSION

As one of the key technology trends within the era of digital transformation, the Internet of Things (IoT) is anticipated
to take immense strides forward within the returning years, with nearly fifty billion devices/things foreseen to be
connected to the web by the year 2020, up from seven billion in 2008.
Using IoT and Machine Learning we can ease the life of customers in the mart. The Application of the Smart Cart
System is not only limited to Automated Billing which could help customers to skip long lines but could also be used
for Product Information and Product Recommendation. These applications can help boost customers' experience in
mart which in return would help Marts to boost Revenue.
The Future Scope of the project could be that this system could also be used for Inventory Management in the marts.
This will help marts to keep track of the products they have left and which products are being more sold or vice versa.

REFERENCES

[1] Nick Statt. (2020, July. 14). Amazon’s new smart shopping cart lets you check out without a cashier [Online].
Available: https://www.theverge.com/2020/7/14/21323421/amazon-dash-cart-smart-grocery-shopping-
woodland-hills-store-cashierless
[2] Superhii, (2020, Sept. 4). Objectives of smart shopping cart are changing [Online]. Available:
https://www.superhii.com/superhi/news_d?NEWS_ID=b061b4e8da134583b3e395ee996bf0a6
[3] MBates, “Arduino Uno,” Jan 2013
https://datasheet.octopart.com/A000066-Arduino-datasheet-38879526.pdf
[4] Trossenrobotics, “FSR 402 Data Sheet,” FSR 402 datasheet, Oct 2010
http://www.trossenrobotics.com/productdocs/2010-10-26-DataSheet-FSR402-Layout2.pdf
[5] Xiamen amotec display, “Specifications of LCD Module”, 16x2 LCD datasheet, Oct 2008,
https://components101.com/displays/16x2-lcd-pinout-datasheet
[6] Joseph Nelson. (2020, March. 16). Getting started with LabelImg for Labeling Object Detection Data [Online].
Available: https://blog.roboflow.com/labelimg/
[7] Manish Gupta. (2020, May. 30). Yolo – You Only Look Once [Online]. Available:
https://towardsdatascience.com/yolo-you-only-look-once-3dbdbb608ec4

IJARESM Publication, India >>>> www.ijaresm.com Page 2584

You might also like