Xtract: Computer Vision Shipping Label Extraction for Supply Chain Optimisation

With the advent of ecommerce, the importance of warehousing and parcel management has grown due to the increase in volume of ecommerce transactions. However, a lot of processes within the warehouses often rely on manual processes to handle inbound and outbound shipments, leading to inaccurate inventory or incorrect shipments that increase operating cost. The problem is especially pronounced for third-party warehouses as cartons and shipping labels often do not follow a standardised format.

Xtract is an AI model which aims to alleviate and improve the productivity of the warehouse workers by providing Optical Character Recognition (OCR) capabilities to read and extract carton and shipping label information. The model allows key data, such as shipping identifiers and lot numbers, to be automatically ingested into a Warehouse Management System (WMS).

Trained on a dataset of images from local warehouses, Xtract is specially tuned by IMDA to work in the supply chain context. Solution providers who integrate Xtract will be able to help their customers achieve fasting processing times, reduced need for manual checks, and an accurate warehouse inventory. Xtract does this by performing more accurate text recognition on the carton labels, while simultaneously scanning any barcodes on the label. With one scan, all the information on the carton label can be mapped to fields within the WMS – enabling the implementation of seamless tallying of the cargo

Xtract is containerised and can be easily deployed. Integration and customisation can be easily achieved by APIs. For example, Xtract can be deployed handheld scanners, or on scanners running on a conveyor belt if the labels are positioned correctly.

Xtract is currently in a closed beta and is available at no cost to all Singapore-based solution providers.

Interested solution providers can request access to the beta at AI@imda.gov.sg.