CheckSmart™ is the world’s first, truly smart self-checkout system for departmental stores and retailers.
On traditional self-checkout terminals in departmental stores, customers have to manually put the cart item in front of barcode reader to scan the identity of the item.
This system causes inconvenience to the customers, and often require time and cause recognition error for some products. CheckSmart self-checkout solution changes the scenario, forever !!
CheckSmart™ brings latest A.I. powered GPU accelerated machine vision based self-checkout system for departmental stores and retailers. The device is a significant improvement over traditional UPC/barcode based self-checkout systems used in retail stores like Walmart (provided by IBM). CheckSmart uses Deep Convolutional Neural Network for object recognition with up-to 99.2% accuracy. Together with digital weighing scale, the system works well for all sorts of objects to be checked out during a typical shopping session on retail stores without any human intervention. The system could potentially replace traditional barcode recognition system to improve checkout time and hassle.
Barcode recognition is not so sophisticated as the user must manually find barcode on product body and hold it to the scanner for identification. This tedious process is time consuming and cumbersome. Computer Vision based checkout system could effectively reduce checkout time with acceptable accuracy and reduced false positives. Recent breakthroughs in image recognition using deep learning made it possible to use as a checkout solution from database/inventory items. Moreover, live video inferencing feature enabled fraud detection pretty accurately.
The powerful mobile GPU embedded on CheckSmart is trained on 150 regular grocery items as a preliminary database. The system can visually recognize any object from database placed upon the platform, measure its weight and count similar objects. Now we need to develop a complete database of more items and increase the accuracy without any false positives. The visual recognition takes approximately less than 10 milliseconds but we are trying to reduce the latency with more powerful GPU.
The fundamental improvement with the CheckSmart self-checkout system is the convenience of use for every customers. The system could also provide a truly smart self-checkout experience like never before. The machine is able to recognize a object rather than recognizing the code labeled on it – that’s truly new and exciting for both shoppers and retailers.
People now use IBM/TOSHIBA provided self-checkout systems on retail stores like Walmart. The systems still use traditional barcode/UPC scanning to recognize objects which poses difficulty to identify certain objects. Moreover, customers not familiar with the system feel uncomfortable to use it. Also, there is no surveillance monitoring system on the terminal for fraud detection. Thus only limited retail stores worldwide use the barcode scanning self-checkout systems.
Other self-checkout system manufacturers still use traditional barcode scanning which is not so convenient for all types of customers without specific instruction. They are mainly focused on improving the steps involved in checkout process rather improving the object recognition system. With the CheckSmart self-checkout system , we have improved the basic object recognition system from live video feed with latest GPU acceleration technologies. Traditional self-checkout system providers still struggle with fraud detection solutions on their terminals. The CheckSmart solution significantly improves real-time fraud detection during a checkout process.
CheckSmart Demo video
As of the end of 2008, there were 92,600 self-checkout units worldwide. The number was estimated to reach 430,000 units by 2014. We want to target the global rising demand of self-checkout systems by 2020 and replacing existing systems. As projected, 225,000+ units demanded worldwide within 2020.