Expedition of Bakery Processes for Increased In-Store Productivity Through a POS Object Detector Module for Pastries

In most bakeries, the bottleneck occurs at the cashier's point-of-sales (POS) station, where order encoding is often time-consuming and requires high human concentration. The need to resort to a manual encoding process results from how pastries are packaged and sold in stores. Since pastries are often freshly baked and sold, they do not have physical price tags. This weakness in an industrialized, fast-paced society is uncommon as POS in other industries uses radio frequency identification (RFID) and barcodes. Such methods are impractical for open bakeries that use unpackaged, freshly baked pastries meant for immediate consumption. The absence of POS order encoding process automation creates a risk of order confusion, resulting in long waiting queues and poor customer service. This study addresses the bottleneck's root cause by developing a Faster R-CNN-based POS module for pastry detection, classification, and receipt generation. This module aims to increase retail store productivity by expediting the POS order encoding process, was validated through a time study.