An Alabama-based tech company has developed an artificial intelligence (AI)-based model that creates recipes according to the contents of a user’s refrigerator.

Researchers at PeopleTec reportedly combined raw food and recipe ingredients via image analysis and a language model to create a realistic recipe, complete with the title of the recipe as well as the proportion and steps necessary to create said recipe.

To develop the technology, the computational model was trained on more than 2,000 images of open refrigerators featuring assorted raw food ingredients. From this, the AI model created a 100-page recipe book featuring unique recipes based on these image inputs.

This is accomplished using application programming interfaces (APIs), which can recognize objects, and GPT-4, which is an OpenAI large language model (LLM) text generator, according to its developers.

“To demonstrate the API as a modular alternative, we solve the problem of a user taking a picture of ingredients available in a refrigerator, then generating novel recipe cards tailored to complex constraints on cost, preparation time, dietary restrictions, portion sizes, and multiple meal plans,” the researchers explained.

The AI-recipe generator is detailed in the article, “The Multimodal And Modular Ai Chef: Complex Recipe Generation From Imagery,” which appears in the journal arXiv.

To contact the author of this article, email mdonlon@globalspec.com