AI-Driven Recipe Discovery: A Survey on Ingredient Identification and Recipe Generation from Food Images

Abstract

The advent of artificial intelligence (AI) has revolutionized numerous fields, and the culinary domain is no exception. AI-driven recipe discovery, particularly through the identification of ingredients from food images and the generation of recipe instructions, has the potential to transform the way recipes are created and personalized. This survey explores the state-of-the-art techniques in ingredient identification and recipe generation, focusing on leveraging deep learning models and computer vision for extracting relevant data from food images. We review the various methods employed for ingredient recognition, such as Convolutional Neural Networks (CNNs) and object detection models, alongside the use of sequence modeling techniques like Recurrent Neural Networks (RNNs) and Transformer models for recipe generation. The survey highlights key challenges, such as ingredient ambiguity, contextual understanding in cooking processes, and dataset limitations. Furthermore, it provides an in-depth evaluation of current AI models based on accuracy, creativity, and practicality in recipe generation. Finally, we propose future research directions that aim to enhance the capabilities of AI in culinary applications, emphasizing the integration of multimodal data and the development of more comprehensive, real- world recipe discovery systems. This work aims to contribute to the growing intersection of AI, food science, and culinary innovation.