PACKDIFF
Adaptation of the text recognition model (currently tailored for Latin-based languages) to Semitic languages, with an additional constraint on ligature characters.
The project
Duration: Jan. 2018 – Present
5 Degrés is supporting the IT department and business teams of a major retail company as part of the PackDiff project.
The project aims to support teams in the design, validation, and inspection of product packaging, particularly for private-label products.
Until now, business teams have manually verified that the information on the packaging matches the product data. This process has several limitations: a legal risk in the event of errors in product information, a time-consuming and tedious process, and difficulties in comparing different versions of the packaging.
To address these challenges, 5 Degrés has collaborated with a partner to develop a solution that automates the verification of compliance between packaging and product data. The tool automatically analyzes packaging files (PDFs), extracts textual data and pictograms, and compares this information with the source files to detect any discrepancies.
The system also allows users to compare different packaging versions, helping marketing teams with design updates or changes to product content.
Recent Developments
As part of the modernization of the client’s application ecosystem, several technical upgrades have recently been implemented:
- Gradual integration of PackDiff into the PLM environment
- Migration of legacy application workflows to PLM middleware routes and APIs
- Phased Phase-Out of Codev (SOAP) Calls
- Implementation of SSO authentication to secure access to the tool
- Adapting PackDiff workflows to the new PLM interfaces and endpoints
The result
Automation of a large part of the packaging compliance verification process. The information on packaging can now be automatically extracted, analyzed, and compared with product data, making it possible to quickly identify any discrepancies.
This automation has improved the reliability of the information printed on packaging, reduced the risks associated with product errors, and increased the efficiency of compliance checks.
Coming soon
The next steps in the project will focus on finalizing the integration with the PLM environment, continuing the testing phases, and stabilizing the data exchange flows between the various applications.
Work is also planned to improve the solution’s automatic analysis capabilities by enhancing the text recognition models used.
TOOLS USED
