SMART RECIPE GENERATOR: A DEEP LEARNING-POWERED MOBILE APP FOR PERSONALIZED AND WASTE-REDUCING FOOD
Loading...
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Universiti Malaysia Sarawak (UNIMAS)
Abstract
Description
The"Smart Recipe Generator" is a mobile application powered by deep learning solutions to change meal preparation to one that solves food waste and offers personalized nutrition needs. This project employs artificial intelligence with state-of-the-art image recognition technologies to recognize vegetables and fruits from user-uploaded images. It creates customized recipes via an Gemini API based on available ingredients as well as diets, restrictions, and preferences. Using TensorFlow and an Agile methodology, it ensures development will be iterative and user-centered. People should not throw away food they haven't used very much, rather, they can identify recipes for ingredients about to expire. The system also encompasses various dietary considerations, e.g. vegan, gluten-free, or low-carb diets. This Android solution has sustainability and health dimensions that intend to make life easier for consumers, moving them toward more informed, environmentally friendly food choices. Full application in all its functionalities as well as real-time identification with recipe modifications based on ingredients would add up to resource savings with individualized nutrition control. The application has received a lot of positive reviews during UAT. 80% of testers rated the application very easy to use (registration, login, and built-in function navigation), while 20% found it easy. Regarding performance, 12 of the 15 respondents gave both detection and recipe generation speeds very fast ratings. Functionally speaking, 100% of the users indicated the system did save recipes and pantry ingredients, and 93% said it did generate recipes correctly based on selected preferences. Another 87% strongly agreed on the app going a long way towards lessening food waste and better planning of meals. One of the most critical areas for future work is building on this model to allow for the detection of a wider range of food ingredients beyond the current 120 so as to cater to more diverse meals and cultural food varieties.