The research focuses on the current challenges for product recognition systems in retail, especially for grocery stores. There exists four main challenges today: * Robust recognition: Robustness in this case is defined as a computer vision system that is scale-invariant, handles uneven illumination, motion blur, and variation in camera angle. * Fine-grained recognition: Many products in a grocery store are similar and there are only small subtle differences between them. * Training of a product recognition: The different number of products in a grocery store contains and huge number of classes with an imbalanced distribution. In addition, these models should be able to update in an online manner in an efficient way. * Domain adaptation: this involves methods to transfer models between different environments (they should be somewhat similar) where there is limited or no samples in the target domain. The purpose of this project is to address these challenges and evaluate new models and methods to expand the knowledge of product recognition systems in retail.