Core demands
The BOB SYSTEM equipment installed at the supermarket checkout will collect a large number of shopping cart image data every day, but the system cannot filter invalid pictures without shopping carts, nor can it identify whether there are unsettled goods on the shopping carts, which creates a certain bottleneck for the upgrade of the risk control system.
Solution
In March 2018, after accessing Baidu EasyDL customized image recognition, BOB SYSTEM was able to divide the collected images into three categories in real time: no shopping cart images, empty shopping cart images and images of goods with cars, and display the pictures of goods with cars to the cashier in real time to help the cashier quickly check whether there are any missing unsettled goods.
After Baidu EasyDL was added to the data processing flow, the identification accuracy of the target image was greatly improved, thus improving the accuracy of data analysis, and the inspection experience of the terminal cashier was also improved. The latest statistical results show that after the use of the system, the loss of each cashier is reduced by an average of 53 dollars every month. This case covers 133 supermarkets in total, and each supermarket has about 10 cash registers, which reduces the loss of the group by about 70000 dollars every month.