Description:Title:

AUTOMATED RETAIL CHECKOUT SYSTEM USING YOLOv5 IN COMPUTER VISION

Field of the Invention

[0001] The present invention is related to the computer science domain and computer vision field.
[0002] This innovation refers to an automated checkout system that can generate the bill of the items purchased by the customer. This system can help to improve customer satisfaction by providing a fast and convenient checkout experience. Furthermore, it can also help to reduce labor costs for retailers by automating tasks that would otherwise require staff members or setting up of expensive machines.

Background

[0003] The background description provided here includes all the relevant information that may be beneficial in understanding the present invention. It is not an acceptance that any of the information or a fact provided herein is prior invention or relevant to the currently claimed invention, or that any publication categorically or implicitly referenced is a prior art.
[0004] The pursuit of generating this system was cultivated after comparative observations of the real-life events that are occurring in and around the globe. With the advent of technology, it has been noticed that tasks are now performed without human intervention at a larger scale. In grocery stores, it has been observed that there are longer queues for customers to checkout and pay the bill for the bought items. Besides there has been increased cases of theft and it causes great loss for the retailers. Though security cameras and high-level automated machines that scan the barcode are present in today’s market, these are cost ineffective. Hence there needs to be a system that can easily generate the bill in cost effective way. This system can be helpful not only to the retailers but also to the customers who can easily cross check the products that have been bought.
[0005] The main aim of the invention is to ease the process of checkout at the retail stores at much cheaper costs. It aims to provide smoother checkout experience for both the customers as well as the employees working at the retail stores. Apart from this, this invention targets the reduction in the cost required to manage the retail stores. It can help the retail store owners to reduce the costs of the manual labor required to maintain the checkout corner. Besides, it can help to eliminate the risks of missing any item during the checkout and potential thefts that generally arise at these stores. The retailer will also be able to efficiently check the stocks of the items that can further enhance the management of retail stores.
[0006] There is a scope for enhancement of this invention. As technology evolves rapidly, there can be new advancements that can be integrated in this invention. Retailers might integrate automated checkout using YOLOv5 with mobile apps that can help the customers to scan the items themselves. Instead of depending on the computer or on human, the customers can crosscheck their bought items and only pay for the intended items. With it, they can be able to pay using the digital transactions as well. This can improve the efficiency of the checkout process and the customers do not have to wait in queues for long time.
[0007] Real time processing of the items can also be achieved. The items can be placed on the conveyor belt or the checkout counter and their prices can be calculated in the real time. This can improve the checkout process at much larger scale. Furthermore, there can be advancements made to handle dense checkout areas as well as to provide error free recognition of various item shapes and sizes. Because of YOLOv5's easy to adapt architecture, the retailers can customize the system according to their store layouts and product categories. This can enhance the store’s services and management and can thus help to fulfil the customer demand fruitfully. This system can also have the potential to assist the retailers in tracking the items and managing the stocks as well.

Objects of the Invention

[0008] Following are the objectives of the present disclosure:
• To pre-process the dataset and convert it into required CSV format.
• To prepare YOLOv5 model using Roboflow.
• To empirically evaluate the outcomes of the proposed model.
• To assess the outcomes of the proposed model with previous models.

Summary

[0009] This system can detect multiple objects:
• The end user can upload the video or image of the bought products.
• An automated bill is generated.
• The bill contains the name of the individual product, its quantity, and its price.
• The total amount of the products bought gets calculated.
• The size of the dataset can be increased in order to get higher accuracy.

Drawings

Figure 1: Algorithm Process Flowchart

Brief Description of the Drawing

[0010] The figure 1 represents Algorithm Process Flowchart of the working model in the present invention.
Detailed Description

[0011] The user input the image or video of the item and it is matched with the available items in the dataset. If the item is found then object detection and recognition is done using YOLOv5 computer vision software. Inventory of the item is also updated at the time of check-out and bill generation.
[0012] If the item is not found in the dataset, then user is asked to enter a new image or video of the item to be checked-out or a new item can be entered into the invented system.
[0013] The Technology used in the invention is as follows:
• Programming Language:
? Python
• Toolkit:
? Roboflow
? YOLOv5
• User Interface:
? Tkinter
[0014] Following are the end users of the invention:
• Government and Private Retail Stores
• Supply Chain Management Systems
• Customers or Clients in Retail Industries

Advantages of the Invention

[0015] Following are the advantages of the invention:
• Efficient product recognition as YOLOv5 can recognize multiple products simultaneously and efficiently with high accuracy and speed.
• Establishment of seamless checkout experience for customers by eliminating manual labor and longer queues.
• Reduce in the costs of manual labors as well as expensive machines for retailers.
• Reduced errors in scanning the items and preventing thefts in stores.
• Can track inventory in real-time, helping retailers monitor stock levels and improve supply chain management. , Claims:Following are the claims of the invention:
1. The developed system uses Roboflow toolkit which is a Computer Vision developer framework used for better data collection, pre- processing, and model training techniques.
2. The user can upload a video or an image of the bought products. The system checks whether the items present in the image/video are present in the dataset or not.
3. In pre-processing, there is an inbuilt feature in Roboflow for giving annotations. We put labels on each of the photos. Along with this, Roboflow increases the dataset by rotating every image with different angles. After processing the dataset from Roboflow we get 1 API key that can be used to import data in YOLOv5 for further detection.
4. Object Detection and Object Recognition is performed through YOLOv5 which is trained on COCO Dataset. YOLOv5 is used to achieve better accuracy in object detection. It is trained on custom dataset.
5. The system can then generate an automated bill that includes the information like name of the product, its quantity, and its price.
6. The final amount of all the bought items is also written at the end of the report.