Automated sushi plate counting application based on deep learning technology

After completing their meals at a sushi restaurant, customers often face delays in settling their bills, resulting in a low table turnover rate. To address this issue, this paper integrates deep learning and image recognition technology, specifically YOLOv8, with Android Studio using the Kotlin language to develop a mobile application (App). The ScreenMatch plugin is employed to maintain component layout consistency across different screen resolutions. Furthermore, a web server and database are implemented using Xampp, while Ngrok is used to facilitate internet connectivity. YOLOv8’s recognition results are displayed on a web page in JSON format, with OkHttp used to retrieve the JSON data and display the number, types, and total cost of plates in the App. This App not only reduces plate-counting errors but also expedites the checkout process for staff, eliminating the need for significant investment in restaurant equipment upgrades. To validate the proposed method, 700 images were used as test data for experimental verification. The results demonstrated that the model achieved a precision of 0.995, a recall of 0.994, a mean average precision (mAP@50) of 0.994, and a mAP@50–95 of 0.840, with a processing speed of 0.03 seconds per image.