Decoded: Discovering Culinary Creativity

Meal Snap is a groundbreaking web application designed to solve the all-too-common problem of staring into your fridge and not knowing what to eat. Whether you're too busy to plan meals or simply seeking culinary inspiration, Meal Snap leverages cutting-edge machine learning to transform your daily meal planning into an engaging, futuristic experience.

Meal Snap uses advanced computer vision and machine learning to scan your fridge's contents and generate personalized recipe suggestions. The app employs a YOLO v12 model for precise object detection and classification, enabling it to recognize a variety of food items. Additionally, a large language model (LLM) converts your natural language inputs like "I want Mexican, I'm vegan" into searchable tags, linking you to a curated recipe database.

Built with Next.js and deployed on Vercel, Meal Snap is engineered for performance, scalability, and a smooth user experience.

In today's fast-paced world, many of us face the dilemma of being hungry without a clear idea of what to cook. Uncertainty about what's in your fridge, coupled with evolving dietary preferences, makes meal planning a significant challenge. Our hackathon challenge was to develop an innovative solution that not only identifies available ingredients but also inspires you with creative, delicious recipes.

• Identify Fridge Contents:
  Utilize computer vision techniques (CNN & YOLO v12) to accurately detect and classify food items from a simple photo of your fridge.

• Generate Recipe Suggestions:
  Employ a large language model to transform natural language queries into actionable tags for retrieving relevant recipes from our database.

• Seamless User Experience:
  Build a responsive, user-friendly web app using Next.js and deploy it via Vercel to ensure high availability and performance.

• Pave the Way for Future Culinary Innovations:
  Integrate machine learning and natural language processing to set the stage for smarter, more context-aware cooking assistance.

Meal Snap is a prime example of how machine learning can revolutionize everyday tasks. By training a YOLO v12 model, we achieved high-accuracy object detection, although the initial training phase required significant fine-tuning to overcome inaccuracies. Combining this with the power of a large language model has enabled us to offer truly personalized recipe recommendations.

This fusion of computer vision and NLP not only solves a practical problem but also showcases the potential for future innovations in culinary technology.

• Recipe Accuracy:
  Early iterations of our recipe-matching algorithm were inconsistent, prompting us to refine our approach through iterative improvements.

• Model Training:
  Training the YOLO v12 model presented challenges, particularly with initial inaccuracies that required extensive data preprocessing and tuning.

• Real-Time Processing:
  Ensuring the application could process images quickly and deliver accurate results necessitated careful pipeline optimization.

• Dataset Diversity:
  Managing a varied dataset of food items under different conditions proved complex but essential for robust model performance.

• Web Application:
  A fully functional, intuitive web app that scans fridge contents and provides tailored recipe suggestions.

• Machine Learning Pipeline:
  An integrated system combining YOLO v12 for object detection and a large language model for processing natural language inputs.

Meal Snap is more than just a hackathon project, it’s a glimpse into the future of meal planning. By merging the power of computer vision and natural language processing, we’ve created an application that not only answers the question of “What’s in my fridge?” but also inspires you with fresh, personalized recipe ideas. Join us as we reimagine the way we approach cooking, one snapshot at a time.

Snap a pic, and let Meal Snap take care of the rest!

Slideshow Presentation# Meal-Snap