Three Core AI Technologies of YI Tunnel
 Image Recognition

• Our system recognizes products based on the world’s leading Con olutional Neural Network (CNN) technology
• The machine now is already able to recognize 10000 SKUs, and apart from standard items, it can also recognize nonstandard items such as vegetables, fruits, clothes, and dishes
• The time it takes to recognize each item is 0.02 seconds, and the accuracy is up to 99.7%
• This technology is used for unattended checkout, to enable real self-checkout, and barcodes, cashiers, RFID tags and conventional POS machines will no longer be needed
• Once collected, the image of an item can be used forever

 Semantic Action Recognition

• Combine knowledge from multiple disciplines including image/video processing, computer vision, pattern recognition, statis tical learning, AI, and cognitive science to perform semantic action recognition against customers
• By analyzing the underlying data such as images or videos, extract information related to human movements from them, and judge a customer’s entire shopping behavior, including items already pur chased, items not purchased, pping routes etc.
• Meanwhile, retailers can define what actions constitute theft, so that the system can automatically alarm when there is theft

 Facial Recognition

• The deep learning-based facial recognition technology can accurately recognize the faces of people in a store
• The technology provides capabilities such as face features recognition, location of key points, 1:1 face comparison, and 1:N face recognition
• It can easily recognize all the faces in a static image or video and count them, so as to take statistics of foot traffic in a retail location based on human faces
• It can gather member information through facial recognition to provide personalized services and help to achieve smart customer relationship management (CRM)

Three Core AI Technologies of YI Tunnel
 Image Recognition

• Our system recognizes products based on the world’s leading Con olutional Neural Network (CNN) technology
• The machine now is already able to recognize 10000 SKUs, and apart from standard items, it can also recognize nonstandard items such as vegetables, fruits, clothes, and dishes
• The time it takes to recognize each item is 0.02 seconds, and the accuracy is up to 99.7%
• This technology is used for unattended checkout, to enable real self-checkout, and barcodes, cashiers, RFID tags and conventional POS machines will no longer be needed
• Once collected, the image of an item can be used forever

 Semantic Action Recognition

• Combine knowledge from multiple disciplines including image/video processing, computer vision, pattern recognition, statis tical learning, AI, and cognitive science to perform semantic action recognition against customers
• By analyzing the underlying data such as images or videos, extract information related to human movements from them, and judge a customer’s entire shopping behavior, including items already pur chased, items not purchased, pping routes etc.
• Meanwhile, retailers can define what actions constitute theft, so that the system can automatically alarm when there is theft

 Facial Recognition

• The deep learning-based facial recognition technology can accurately recognize the faces of people in a store
• The technology provides capabilities such as face features recognition, location of key points, 1:1 face comparison, and 1:N face recognition
• It can easily recognize all the faces in a static image or video and count them, so as to take statistics of foot traffic in a retail location based on human faces
• It can gather member information through facial recognition to provide personalized services and help to achieve smart customer relationship management (CRM)