The ProdID EU Project Comes to an End: Advancing in Smart Product Identification for Industry and Commerce

The ProdID EU project, conducted at the Computer Vision Centre (CVC), has successfully developed multimodal image understanding technologies for fine-grained, real-time product identification in industry and commerce. This initiative has integrated visual and textual data to enhance product retrieval, addressing key challenges in inventory management, automated check-out systems, and visual product search.

Key Technological Advancements

1. Multimodal Product Retrieval – Combining visual features and recognized text from product packaging to enable accurate identification, particularly for visually similar products.
2. Regular Expression-Based Text Spotting – A novel method allowing structured text searches (e.g., serial numbers, prices) for industrial applications.
3. Synthetic Data Generation for Retail Environments – Creating virtual supermarket shelves with labeled products to enhance training of AI models.
4. Zero-Shot Learning for Product Matching – Implementing zero-shot retrieval strategies, enabling recognition of unseen products without extensive labeled data.

Industry Applications & Prototypes

As part of the project, three prototypes were developed and tested in real-world scenarios:

• Supermarket Product Retrieval System: A real-time retrieval system that enables product identification using image and text cues, tested in collaboration with Spanish and German companies.
• Library Book Inventory System: A prototype tested in the Miquel Batllori Library (Sant Cugat), allowing for automated book location and inventory tracking using AI-driven retrieval.
• Pharmaceutical Stock Monitoring: A system designed for real-time product tracking in pharmacies and medical stockrooms, with pilot studies conducted with a multinational medical device company.

Market Exploration & Technology Transfer

Throughout the project, market research and industry outreach were prioritized:

• Engagement with 5 key industry partners, including Logista Libros, BBraun, and Autonomo, for potential real-world deployments.
• Presentation at major industry events, including Mobile World Congress, IOT Solutions World Congress, and EXPO FoodTech 2023, to showcase technology applications.
• Patent and Intellectual Property Strategy: Two i-DEPOT patents were secured for image-text matching for book collections and book identification methodologies.

Impact & Future Directions

• Technology Readiness Level (TRL) Progression: The project started at TRL 3-4 and successfully reached TRL 5-6, with prototypes tested in operational environments.
• Time to Market: The technology is expected to be commercialized within 1-1.5 years, with licensing opportunities being explored through CVC’s technology transfer office.
• Potential Cost Savings & Efficiency Gains: AI-driven automated inventory management could reduce operational costs by up to 20%, benefiting industries such as logistics, retail, and healthcare.

ProdID has successfully bridged the gap between research and commercial viability, laying the foundation for intelligent product recognition solutions that can streamline operations, reduce costs, and enhance consumer experiences.

For further details, visit CVC’s ProID project website or contact our technology transfer office.

.

_{Proyecto PDC2021-121512-I00 financiado por MCIN/AEI/10.13039/501100011033 y por la Unión Europea NextGenerationEU/ PRTR}