IMPRO

Today, retail companies often pursue an omni-channel retailing strategy that supports all sales channels such as stores, online stores or catalogs equally. This requires all sales channels to be operated, controlled, evaluated and optimized in parallel. As a result, the amount of information to be managed per product has increased considerably in recent years. This places high demands on both product information management and master data management in the central ERP systems.
Due to e-commerce, product images in particular have become increasingly important. For companies, this means that product data is very extensive and heterogeneous. They consist of structured and unstructured data, as well as image and video data. In many cases, the data still has to be captured manually, which is very cost-intensive and error-prone. It is therefore difficult for companies to ensure that the product data and the information they represent, which come from different sources, are correct.  The reconciliation of the information contained on the product image with the product data records provided, which is required for this purpose, is carried out manually, if at all, and is correspondingly time-consuming.

The aim of the research project is to automate the creation process of product data by incorporating product images and further information from web sources or internal data sources by means of an Image supported Product Data Creation Processor (IMPRO) using machine learning methods. This is intended to reduce costs, especially for the acquisition of product data, and to increase the efficiency and quality of the required maintenance processes.

To Project Homepage