ARTIFICIAL INTELLIGENCE
Automated shelf product recognition with AI
Artificial Intelligence (AI) plays a key role in transforming the retail sector by automating processes that previously required significant manual effort. Thanks to advanced image recognition algorithms, it is now possible to effectively monitor product displays and verify planogram compliance, resulting in greater consistency and operational efficiency.
As part of a collaboration with a leading fuel retail distributor, ALTEN Polska designed and implemented an AI-powered solution for automatic verification of shelf product placement at fuel stations. Based on photos taken on-site, AI models were trained and fine-tuned to recognize products in challenging conditions — including glass, reflections, and non-standard layouts.
Challenge
Automate the process of identifying and verifying product placement in visually complex conditions (reflections, glass, labels, unusual shelving).
Solution
Deploy highly accurate AI models capable of analyzing images from fuel stations, recognizing products, and comparing their arrangement with planograms.
Benefits
- 99.7% accuracy in recognizing 37 confectionery products
- 100% accuracy in identifying 14 cigarette brands
- 99.1% accuracy in detecting baskets and their contents
- Significant reduction of manual inspections
- Improved planogram compliance
- Standardized product displays across the entire network
- Streamlined inventory management
Challenges and Impact on End Users
The project addressed the growing demand for automation in fuel retail, where correct product placement directly affects sales. The complexity of the environment — refrigerator glass reflections, large labels, glass bottles, and irregular shelving layouts — required the creation of exceptionally precise and robust AI models. Ensuring high accuracy was critical so that station staff could rely on the analysis results without manual double-checking, which directly improved their workflow and the overall customer experience.
Implemented Tools and Technologies
The project, led by ALTEN Polska, used a set of modern technologies: Google Cloud Platform, Python and Kotlin programming languages, and the CVAT tool for annotating training data. The AI models were fine-tuned to the specific environmental conditions of fuel stations, enabling adaptation to various display scenarios. The entire process — from image processing to compliance analysis — was fully automated and integrated with the client’s existing systems.
Benefits for the Client and End Users
Thanks to the implemented solution, the client gained not only a highly accurate and reliable system for monitoring planograms but also tangible operational savings. Automated image analysis significantly reduced the time needed for product placement checks, minimized human errors, and improved display consistency across the entire network. Station staff received a tool to support their daily tasks, while end customers benefited from a clearer and more organized product presentation, improving the shopping experience.
