SmartPlate is an end-to-end Computer Vision application designed to automate dietary tracking. Unlike generic calorie counters, this project utilizes Transfer Learning on the YOLOv8 architecture to specifically detect and recognize Indian cuisine, mapping detected items to a custom nutritional database in real-time.
The system detects the food item (e.g., Avial) and retrieves the corresponding macronutrients.
A pandas-based breakdown of Calories, Protein, Carbs, and Fats.
- Custom Object Detection: Fine-tuned YOLOv8 model trained on a custom dataset of Indian foods.
- Intelligent Mapping: Robust string-matching logic to map detected classes (CV output) to a nutritional CSV database.
- Interactive Web App: User-friendly interface built with Streamlit allowing image uploads and instant analysis.
- Error Handling: "Graceful failure" mechanisms that guide users when non-food items or unknown dishes are detected.
- Deep Learning: YOLOv8 (Ultralytics), PyTorch
- Web Framework: Streamlit
- Data Manipulation: Pandas, NumPy
- Image Processing: PIL (Python Imaging Library), OpenCV
- Version Control: Git & GitHub
AI Nutrition Assistant/
├── datasets/ # Dataset directory
│ └── food_dataset/ # Custom Indian food images & labels
├── app.py # Main Streamlit application
├── nutrition_data.py # Logic for CSV parsing and data retrieval
├── nutrition_info.csv # Database of food items and macros
├── train_model.py # Script for training/fine-tuning YOLOv8
├── requirements.txt # Project dependencies
└── README.md # Documentation
Arijeet Dutta Engineering Graduate Email : arijeetdutta501@gmail.com LinkedIn