TamaLife is a modern take on the classic Tamagotchi virtual pet, combining mobile app technology, machine learning, and IoT hardware to create an interactive pet care experience. Built during CalgaryHacks 2025.

• Virtual Pet Care: Feed and interact with your virtual pet through a mobile app
• ML-Powered Food Detection: Take photos of food to feed your pet, with real-time classification of food types
• Hardware Integration: Physical LCD display and LED indicators showing your pet's mood and health
• Natural Language Interaction: Talk to your pet using advanced language processing
• Health & Mood System: Dynamic pet state that responds to your interactions and care

• React Native with Expo
• TypeScript
• TensorFlow.js for ML model integration
• Custom animations and UI components

• Node.js with Express
• MongoDB for data persistence
• JWT authentication
• WebSocket for real-time updates

• TensorFlow/Keras for food classification
• Custom trained model on food/non-food dataset
• Model conversion to TFLite for mobile deployment

• Arduino microcontroller
• LCD display
• RGB LED indicators
• Serial communication with backend

1. Clone the repository:

git clone https://github.com/yourusername/tamalife.git

2. Install dependencies:

# Root directory
npm install

# Backend setup
cd backend
npm install

# ML environment setup
cd ml
python -m venv venv
source venv/bin/activate  # or `venv\Scripts\activate` on Windows
pip install -r requirements.txt

3. Set up environment variables:

# Create .env files in both root and backend directories
cp .env.example .env

4. Start the development servers:

# Backend
cd backend
npm run dev

# Mobile app
npm start

The project utilizes two major datasets for training the food classification model:

• 101 food categories
• 1000 images per class
• Used for training food/junk food classification
• High-quality, real-world food images
• Source: Food-101

• Binary classification dataset (food/non-food)
• 2500 food images
• 2500 non-food images
• Training, validation, and evaluation splits
• Source: Food-5K

The datasets were combined and reorganized into three categories:

• Non-food (from Food-5K)
• Healthy food (selected from Food-101)
• Junk food (selected from Food-101)

Data split:

• Training: 70%
• Validation: 15%
• Testing: 15%

Ensure proper serial port configuration in backend settings

Multiplayer pet interaction More sophisticated pet behavior patterns Extended food classification categories Enhanced hardware display capabilities Voice command integration