Cashier Fraud Monitoring
Automatic detection and notification of fraudulent acts based on camera images at the top of the self-checkout machine, thereby contributing to the protection of store profits and the elimination of the shortage of monitoring personnel.
Challenges
Carelessness or intentional misconduct in self-checkout machine resulting in increased losses.
Lack of manpower and difficulty allocating sufficient staff for monitoring.
Solutions
Automatically detect and notify a fraudulent act in a self-checkout machine by recognizing an item or a person based on image analysis.
Functions
・Scan avoidance detection(Features in development that are not installed in Kozuchi)
・Detection of disguise of scanning items (Banana trick and label switches)
・Detection of remained unscanned items in baskets.
Strength
World’s most accurate detection of “An object held in hand”. (2023)
*Detection accuracy for “An object held in hand” on HICO-DET dataset (presented in VISAPP 2023)Winner of competition of product recognition at self-checkout on international conference. (2023)
*CVPR AI City Challenge 2023
Usage Image
Concept Video
Demo Video
Document
Not for corporate use