Automated Recipe Generation using Ingredient Classification based on an Image from a Real-Time Photo Station

This paper presents a novel solution to address the challenge of recipe selection based on available ingredients in a household, particularly for new cooks or even experienced chefs. Leveraging the power of technology, specifically machine learning, this study introduces a "recipes as a service" concept that utilizes object recognition through image processing. By taking a single photograph of the ingredients on a kitchen counter or refrigerator in real-time, the system generates a list of all possible recipes that can be made from the identified ingredients, enabling users to maximize their kitchen innovation. The study evaluates several image classification and correlation models, including Efficient Net-Lite, faster-RCNN, YOLOv4, and YOLOv5, to identify the best model for the image recognition tasks. The comparison is based on various metrics, including accuracy and efficiency, and the results show that YOLOv5 is the optimal model for the purpose. The proposed solution provides an automated recipe generation system that can help users overcome the challenge of selecting recipes and planning meals daily. The system can be operated in real-time, making it a valuable tool for households. The results of the study can potentially contribute to the development of smart kitchens and future innovations in the field of culinary technology.