The Smart Kiosks From SandStar Demonstrate 'Dynamic Vision Recognition' is for the Future

Which kind of recognition technology is better? Different suppliers of smart kiosks may choose a different technology for a reason. Readers will have a better idea of why so many suppliers prefer dynamic vision recognition after reading this article.


A smart kiosk is really “smart”:  Shoppers swipe their credit card to open the door, grab whatever they want, close the door, and off they go, as the payment is made automatically.

There's no need to press a button to select an item or any stooping down to pick up the item. Shoppers just need to "grab and go." This shopping experience made possible by smart kiosks is appealing to consumers. It has also sparked a battle among kiosk suppliers adopting different recognition technologies.

Smart kiosks today are mainly based on four recognition technologies: weight sensors, RFID tags, static vision recognition, and dynamic vision recognition. They recognize items differently and can all be used for “automatic checkout.”

Which one is better? Different suppliers of smart kiosks may choose a different technology for a reason. Readers will have a better idea of why so many suppliers prefer dynamic vision recognition after reading this article.



Old and troublesome

RFID is perhaps the oldest one among the four. A unique tag is put on each item. When an item is taken out, the RF signal of its tag will be lost and the system will know which item is taken out and its price.

This method is pretty simple. Its shortcomings are obvious, too.

Although it has been widely used for more than 10 years, RFID still can't be used for metal due to signal interference. This means that canned drinks sold in most vending machines can’t be sold in a kiosk using RFID tags. This will limit the items that can be sold and lower sales by 25%.

RFID is quite expensive, too. The tags will be taken away by customers with the purchases. Besides, retailers need to hire people to manually attach the tags. Each tag costs about 10 cents for material and 10 cents for labor. Assuming that 25 items can be sold each day in a single kiosk, retailers will need to spend more than $1,825 a year for tags and labor, which is almost equivalent to the cost of the kiosk itself.

And RFID tags are easy to come off, and the damage rate is up to 10%.


Weight sensors

Higher costs and more limitations

Weight sensors are installed on each layer of kiosk shelves. After an item is taken away, the system will immediately calculate the weight loss against the original weight in the database, to find out which item is taken away.

However, the biggest problem of weight sensors is that all the items need to be put on specific spots and in a specific order, and if one item is mistakenly put by a customer on the spot of another item, the system will be confused and not able to recognize it correctly.

That’s why the weight sensor solution has to work with other technologies to ensure correct recognition in the scenario above, which happens all the time and is not avoidable.

In practice, weight sensor kiosks also require a completely flat ground in order to ensure accurate recognition, which will require frequent recalibration.


Static vision recognition

Too many restrictions

Static vision recognition works by comparing images. The system knows which items are missing by comparing the photos taken before the door is opened and after the door is closed, and adds the missing items to the shopping cart for the buyer.

This solution can’t be widely used because of the restrictions on how the items are arranged. The planogram needs to ensure a good vision for the camera, so items can’t be too tall, and taller ones have to be put to the side and shorter ones in the middle.

This will increase the costs of model training and merchandise monitoring. And customer behaviors are totally unpredictable, so when an item is put on a wrong spot, it may block the vision of the camera, leading to inaccurate recognition and bad user experience.

Kiosks using static vision recognition usually have a camera on every shelf. A fridge has no more than five shelves, as the cameras will take a lot of space, and almost 40% of the space inside will be wasted. Retailers will have to increase the restocking frequency, and the costs will go up by 40%.

In addition, the cost of network connection is very high, too, because with static vision recognition, retailers need to upload all the images to the cloud, where bandwidth requirements are hard to meet in most countries and regions of the world.


Dynamic vision recognition is the future

Today, many retailers, including three of the world’s top three beverage suppliers, have all switched to SandStar’s dynamic vision recognition technology.

And SandStar has also developed the world's fastest and most accurate Computer Vision software system in the world, which not just recognizes the items taken, but also recognizes how they are taken after the door is opened. As a result, its accuracy is much higher than that of other recognition technologies.

It delivers the real-time shopping cart on the smart kiosk or on the handheld device, empowering immediate review and approval of the recognition accuracy by the customer, which can promote consumers’ confidence and enable a frictionless retail shopping experience.

SandStar acts on CSR initiatives - the solution is carefully designed to protect customer privacy. The camera involved only captures products and the shopping behavior by marking the key point of the human skeleton. Additionally, since the system is based on self-learning, all data processing is done automatically by the system itself.

And dynamic vision recognition has no restriction on the types of merchandise and their arrangements, can make full use of the space, can increase efficiency, and can drive up sales by up to 30%.

In addition, the data of customer behaviors collected based on dynamic vision recognition can provide insights for merchandise selection and pricing to help increase sales by 20%.

And also, each time after the shopping journey, the smart kiosk will activate the Blue Light virus cleaning cycle.

In comparison, RFID involves a high damage rate of 10% and reduces sales by more than 25%; weight sensors have to work with other technologies; static vision recognition wastes 40% of the space, requires a higher restocking frequency, and increases costs by 40%, so the dynamic vision recognition solution of SandStar is obviously much better.      

Thanks to four years of hard work by 220 engineers from SandStar, the company’s smart kiosks have gone through eight generations. In this process, SandStar has seen many companies fail. Today, SandStar has the world’s only commercially available smart kiosks, adopted by three of the world’s top beverage suppliers. The company has many Fortune 500 companies using its smart kiosks, including Wanda Group and Sinopec.

SandStar is expanding globally quickly and has set up branches in the United States, China, and Thailand. The Smart Kiosks have been deployed in 12 countries and over 20,000 systems are scheduled to be deployed over the next 24 months as a result of the COVID-19, as retailers' interests in frictionless shopping have increased due to the lower efficiency of the human hand. Both the retailers and the customers are seeking out un-attended check out as their preferred shopping experience. The Smart Kiosks of SandStar have shown they can save time and promote confidence in a healthy shopping experience.

Source: SandStar


Categories: Retail Technology, Retail, Restaurants, Bars, and Catering

Tags: computer vision, intel, IoT, SandStar, Smart Kiosk

Related Video


View Website

SandStar focuses on AI computer vision in the retail scene, using AI technology to constantly explore the edge of new retail. Its business launched in 11 countries, serving 20+ Fortune 500 companies globally, building the AI retail ecosystem.