Calorie burn prediction & fitness analytics
Estimate calories from heart rate, duration and body measurements, keep a private workout history, and see your trends — with every number labeled by the method that produced it.
Free account · Estimates only, not medical measurements
Today
Illustrative preview
Estimated burn
418kcal
Sessions
5
Minutes
184
Goal
78%
Recent workouts
Running
Demo estimate
418 kcal
Cycling
Trained model
352 kcal
Walking
Manual entry
160 kcal
Estimate a workout
Measured inputs in. Labeled estimate out.
Result · Demo Estimate
419.4 kcal
A published heart-rate equation (Keytel et al., 2005) gives a Demo Estimate out of the box. Import a model trained by the included Python pipeline and results become ML Predictions.
Predictions, manual entries and optional sample data live in one history — each tagged so you never mistake one for another.
Age, sex, height, weight, duration, average heart rate and body temperature — each checked against realistic ranges.
The app uses a connected trained model if available; otherwise the Demo Estimate formula. The label tells you which.
Saved sessions flow into your dashboard, analytics, comparisons and reports immediately.
Validated inputs — age, sex, height, weight, duration, heart rate and body temperature — produce a deterministic estimate in kcal.
Add, edit, delete, search, filter and export every session to CSV. Each record keeps its source.
Daily, weekly and monthly trends, duration over time and workout-type distribution from your saved data.
Put two sessions side by side: duration, calories, heart rate and kcal per minute.
Pick a period, print or save as PDF, or download a plain-text summary.
Every workout belongs to your account. Database policies stop anyone else reading it.
Charts are built only from workouts you have saved. With too little data you see an empty state — never invented insights.
Weekly calories
Illustrative preview
Sessions
Your data
Total kcal
Your data
Avg / session
Your data
An included Python pipeline trains and compares three regressors with cross-validation on a public exercise dataset. Metrics appear in the app only after you run training and import the results — none are shown here because none are invented.
A weighted sum of features. Runs directly in the browser once its coefficients are imported.
Averages many decision trees to capture non-linear effects. Served by the optional local API.
Trees built in sequence, each correcting the last. Often strongest on tabular data.
Methodology
Keytel et al. (2005) heart-rate equation, MET fallback.
Always available. Not machine learning.
Methodology
Linear Regression exported by ml/train.py.
Used after you import model_export.json.
Methodology
Best model by cross-validated RMSE, via local Flask API.
Used when you connect the API.