Viscovery Developed AI-Powered Recognition Technology for Bakery Checkout
September 3, 2019
Commercial Times 鄭淑芳 (News Coverage)
Not only can you use AI to check out at convenience stores, but you can also use AI to check out bread which is difficult to identify by the “look”. Viscovery recently announced that its bakery self-checkout system, with AI image recognition technology, has been deployed in a well-known coffee bakery chain, 85°C Bakery Café, in the United States. It is the first AI-powered bakery self-checkout system in the US.
Viscovery bakery self-checkout system uses Deep Learning and Computer Vision technology to replace human eyes to identify bread for checkout. By adding a simple 1080p camera, stores can start to use the AI system. Moreover, the AI system is compatible with almost all kinds of POS system, so the overall setup fee is relatively cost-effective.
With the AI self-checkout system, the computer will automatically identify the bread items that shoppers put on the checkout counter, and identify the quantity, all within 1 second. For consumers, the entire checkout process is fast and convenient, saving abundant waiting time in line and improve the shopping experience; for bakery store owners, the bakery self-checkout system can enhance the checkout efficiency, and further increase revenue, without adding any extra manpower.
The bakery recognition checkout system can be deployed in both self-checkout kiosk systems and POS systems, assisting cashiers to identify bread and check out faster than ever.
Viscovery’s CEO, Kevin Chen, pointed out that the most difficult part of bread recognition is its variance appearance. On one hand, even the same kind of bread may have differences in appearance. For example, the baked color may vary drastically, or the ingredients sprinkled on the bread may be different in terms of the amount and the position. On the other hand, different types of bread may look alike.
Other challenges lie in the bread overlapping scenario, and changes in light and shadow in the recognition area. Although shoppers usually do not stack bread together for checkout, there are chances that partial overlapping or lighting changes may happen from time to time.
Viscovery self-developed model training technology to overcome these challenges, strengthen the detection and recognition capabilities, and even keep the system automatically optimizing and improving the recognition model.
What’s more, to increase the operation flexibility, Viscovery also developed product management system, which is convenient for store owners to add, update or delete product images and keep the database updated. The system can be managed and updated by the headquarters or by each branch.