I-POST: Intelligent Point of Sale and Transaction System

TL;DR

I-POST is an automated, intelligent point-of-sale system utilizing machine learning and facial recognition to streamline shopping by enabling users to walk in, collect items, and exit without queuing.

Contribution

The paper introduces a novel POS system that leverages CNN-based object detection and facial recognition for real-time, automated checkout, improving efficiency over traditional systems.

Findings

Object recognition accuracy of 97%

System enables queue-free shopping experience

Potential for broad application in retail environments

Abstract

We propose a novel solution for the cashier problem. Current cashier system/Point of Sale (POS) terminals can be inefficient, cumbersome and time-consuming for the users. There is a need for a solution dependent on modern technology and ubiquitous computing resources. We present I-POST (Intelligent Point of Sale and Transaction) as a software system that uses smart devices, mobile phone and state of the art machine learning algorithms to process the user transactions in automated and real time manner. I-POST is an automated checkout system that allows the user to walk in a store, collect his items and exit the store. There is no need to stand and wait in a queue. The system uses object detection and facial recognition algorithm to process the authentication of the client and the state of the object. At point of exit, the classifier sends the data to the backend server which execute the payments. The system uses Convolution Neural Network (CNN) for the image recognition and processing. CNN is a supervised learning model that has found major application in pattern recognition problem. The current implementation uses two classifiers that work intrinsically to authenticate the user and track the items. The model accuracy for object recognition is 97%, the loss is 9.3%. We expect that such systems can bring efficiency to the market and has the potential for broad and diverse applications.