Enhancing Retail Checkout with Meta-Learning and OCR: A Streamlined Automation Framework

In recent times, retail automation has gained significant attention, particularly in optimizing checkout processes. Traditional product detection and pricing methods rely on extensive labeled datasets, making them resource-intensive. This paper explores meta-learning techniques to address these limitations, enabling few-shot learning for rapid product adaptation with minimal data. We utilize Prototypical Networks, Siamese Networks, and Model-Agnostic Meta-Learning (MAML) to detect products from camera-captured images and integrate Optical Character Recognition (OCR) for precise brand identification and pricing. To demonstrate feasibility, we present a Streamlit-based application that seamlessly integrates product detection, OCR, and a virtual cart for automated checkout. Our findings highlight the potential of meta-learning in enhancing retail automation, emphasizing scalability, efficiency, and user engagement.