This project integrates generative AI, object detection, and Raspberry Pi to create an automated kitchen capable of preparing a dish based on user input. Leveraging a dataset with over 6,500 recipes, the system guides users from recipe selection to cooking the meal.
The aim is to develop an application that can prepare a dish by following a user's input, automating the entire cooking process.
We utilized the 6000+ Recipe Dataset which allows users to enter keywords or partial titles to retrieve a recipe of their choice.
- Data Conversion: The dataset was converted into a JSON format, resulting in the file processed_recipes.json.
- Search Functionality: Users can enter a keyword or partial title to search for a dish. A dropdown menu displays matching recipes, from which users can select one to view detailed information, including the title, ingredients, cooking time, prep time, and instructions.
- JSON Generation: After selecting a recipe, users can generate a JSON file, actions.json. Google's LLM Gemini is employed to convert the recipe into detailed, step-by-step instructions that are machine-operable.
- Simulation Creation: The generated JSON is used to simulate the cooking process, guiding the flow of the machine.
- Dataset_sources : Spices Dataset Vegetables dataset
- After receiving the ingredients from actions.json the trained model searches for matching ingredients in the conveyor, which is then passed to the required vessel.
- Execution of Instructions: The instructions in actions.json are used to execute cooking steps.
- Raspberry Pi 4: A Raspberry Pi 4 with GPIO is employed to control the execution. GPIO pins are assigned for each step, enabling the automation of various cooking tasks.
- Recipe Selection Code
- Processed Recipes JSON
- Simulation Code
- Object Detection Code
- Hardware Control Code
This comprehensive approach ensures a seamless experience from selecting a recipe to preparing a dish, making cooking an automated and effortless task.