One of the goals of Esko is to create/offer solutions that help to automate the creation of packaging artwork (labels and packages) based on content (claims, ingredients, nutrition facts) stored in a content management system.
However the brand owners and studio's they work with to create packaging artwork (graphics) often do not have the source content available in a structured way today. Often, they rely on approved existing versions of the artwork as single source of truth for the latest approved version of the customer-facing content. To facilitate migration to a more automated workflow to update artwork based on structured content, this project aims to develop a tool that can:
1. Extract textual information from an artwork file.
2. Interpret the extracted text, i.e. recognize text groups, and recognize the purpose of these text groups.
(e.g. whether the extracted text is an address, nutritional information,...).
3. Offer a UI to review the extracted and interpreted text.
Research what are the best techniques to interpret the extracted text data, can we utilize neural networks for this or can this be outperformed by a self-made heuristic algorithm.

Tags: Neural networks, AI, OCR/Segmentation

Java