developerPratik/vitamind-calorie-predictor

🏃‍♂️ VitaMind AI - Calorie Goal Predictor

Personalized daily calorie recommendations using AI

Model Description

VitaMind AI predicts optimal daily calorie intake based on:

• Demographics (age, gender, height, weight)
• Activity metrics (steps, heart rate)
• Lifestyle factors (sleep, stress, mood)
• Activity level (sedentary to athlete)

Performance

• MAE: 75 kcal
• RMSE: 95 kcal
• MAPE: 3.2%
• R² Score: 0.89

Quick Start

from huggingface_hub import hf_hub_download
import tensorflow as tf
import joblib
import numpy as np

# Download model
model = tf.keras.models.load_model(
    hf_hub_download(repo_id="developerPratik/vitamind-calorie-predictor", filename="model.keras")
)
scaler = joblib.load(
    hf_hub_download(repo_id="developerPratik/vitamind-calorie-predictor", filename="scaler.joblib")
)
encoders = joblib.load(
    hf_hub_download(repo_id="developerPratik/vitamind-calorie-predictor", filename="encoders.joblib")
)

# Example prediction
user_data = {
    'age': 30, 'weight': 75, 'height': 175, 'steps': 8000,
    'heart_rate': 72, 'sleep_hours': 7.5, 'stress_level': 4,
    'activity_level': 'Active', 'gender': 'M', 'mood': 'happy'
}

# Feature engineering
bmi = user_data['weight'] / ((user_data['height'] / 100) ** 2)
good_sleep = 1 if user_data['sleep_hours'] >= 7 else 0
high_stress = 1 if user_data['stress_level'] >= 7 else 0
activity_scores = {'Sedentary': 1, 'Lightly Active': 2, 'Active': 3, 'Very Active': 4, 'Athlete': 5}

# Encode
activity_encoded = encoders['activity_level'].transform([user_data['activity_level']])[0]
gender_encoded = encoders['gender'].transform([user_data['gender']])[0]
mood_encoded = encoders['mood'].transform([user_data['mood']])[0]

# Create feature vector (14 features)
features = np.array([[
    user_data['age'], user_data['weight'], user_data['height'],
    user_data['steps'], user_data['heart_rate'], user_data['sleep_hours'],
    user_data['stress_level'], bmi, activity_encoded, gender_encoded,
    mood_encoded, good_sleep, high_stress, activity_scores[user_data['activity_level']]
]])

# Predict
features_scaled = scaler.transform(features)
calories = model.predict(features_scaled, verbose=0)[0][0]
print(f"Recommended daily calories: {calories:.0f} kcal")

Model Architecture

Input (14 features)
    ↓
Dense(256) + BatchNorm + Dropout(0.3)
    ↓
Dense(128) + BatchNorm + Dropout(0.3)  [Residual Connection]
    ↓
Dense(128) + BatchNorm + Dropout(0.3)
    ↓
Dense(64) + BatchNorm + Dropout(0.2)
    ↓
Output (1 - calories)

Total Parameters: ~85,000

Features

Feature	Type	Description
age	int	Age in years (18-100)
weight	float	Weight in kg (40-150)
height	float	Height in cm (140-220)
steps	int	Daily steps (0-30000)
heart_rate	int	Resting heart rate (50-120)
sleep_hours	float	Hours of sleep (3-12)
stress_level	int	Stress rating (1-10)
bmi	float	Calculated BMI
activity_level	str	Sedentary/Lightly Active/Active/Very Active/Athlete
gender	str	M/F
mood	str	happy/neutral/sad/anxious
good_sleep	binary	1 if sleep >= 7 hours
high_stress	binary	1 if stress >= 7
activity_score	int	1-5 based on activity level

Limitations

⚠️ Important Disclaimers:

• For educational/wellness purposes only
• NOT a substitute for professional medical advice
• Individual metabolism varies significantly
• Does not account for medical conditions
• Consult healthcare providers for medical decisions

Training Details

• Framework: TensorFlow 2.15
• Training samples: 5,000 synthetic
• Validation split: 15%
• Test split: 15%
• Optimizer: Adam (lr=0.001 with ReduceLROnPlateau)
• Loss: MSE
• Regularization: L2 (0.001) + Dropout + BatchNorm
• Early stopping: Patience=30

License

MIT License - Free for commercial and personal use

Citation

@software{vitamind_ai_2025,
  author = {{Your Name}},
  title = {{VitaMind AI Calorie Predictor}},
  year = {2025},
  publisher = {Hugging Face},
  url = {{https://huggingface.co/developerPratik/vitamind-calorie-predictor}}
}

Contact

• Issues: Open an issue on this model's discussion page
• Email: [email protected]

---

Built with ❤️ using TensorFlow and scikit-learn