Automatic visual inspection and validation


Computer vision technology designed for high-level understanding of images or video frames, for quick, automated and recurrent auditing tasks, powered by artificial intelligence and business logic.



Auditing in retail or warehousing is still a very time-consuming and manual process, being very prone to human error, while performed by visual inspection and mapping with a plan. Planning layouts are crucial for space optimisation and lean operation. The proof of compliance with store and shelf layouts is a recurrent activity. Wrong product placement or out-of-stock scenarios induce revenue shrinking, contractual penalties with suppliers and poor consumer satisfaction. Quality control in manufacturing is often performed by manual inspection to ensure that parts are assembled short of imperfections. Quality checks are also part of the supply chain and logistics, including the detection of counterfeit products which is a concern increasingly prominent for global brands.


Computer vision and cutting-edge machine learning are used to automate visual inspection, leveraging: object recognition based on image features or context; image labeling; sensor fusion for depth detection and panorama stitching; text recognition and barcode code scanning. For the food retail market, we created a semi-automatic buggy to take high quality images of the supermarket shelves in fast motion, controlling reflections and blur. Shelf layouts are automatically compared with actual in-store products, enabling the detection of wrong placements, out-of-stocks and incorrect prices. User interfaces include highly visual information of issues and priorisation of corrections, with fully integration with information systems by web services. For counterfeit detection, we developed a simple mobile application that guides the user to take pictures of relevant regions of products, extracts information from the acquired data and validates it with databases of genuine products.


More than 200 audits in retail aisles have significantly decreased wrong product placement and out-of-stock, while reducing costs of control. Pilots in the biggest retailers in Portugal and Austria have shown a 90% reduction of monitoring time when compared with human auditing. The counterfeit detector has shown high reliability regarding information extracted from smartphone pictures, of power tools, their packages and nameplates.

Highlighted Projects



In the ShopView project we created a system which uses state of the art image
processing, to automate the task of
controlling the implementation of planograms in retail spaces.



AICOS will address issues identified in ShopView, and complete the validation of the solution throughout a long-term pilot, with demonstrators all over the world, through the ShopView2Market project.



In order to address the counterfeiting, we built a solution based on a mobile application, that communicates with a back-end server for validating the authenticity of equipment being analysed.

Further information




Audit Brochure


Relevant Services


Rapid Prototyping

Innovation Studies

Education & Training


Relevant Publications


Rosado L., Gonçalves J., Costa J., Ribeiro D., & Soares F. (2016). Supervised learning for Out-of-Stock detection in panoramas of retail shelves. In Proceedings of IEEE International Conference on Imaging Systems and Techniques (IST), 406-411. More info

Gonçalves J., Ribeiro D., & Soares F. (2015). Perspective Correction of Panoramic Images created by Parallel Motion Stitching. In Proceedings of 23rd International Conference on Computer Graphics, Visualization and Computer Vision, 125-132.