BodyMetrics Regression System is an advanced AI-driven application designed to model and predict the complex relationships between human age, height, and weight, now featuring gender-specific biometric modeling. Unlike standard linear models, this system uses a multi-directional neural network to estimate missing physical metrics based on available data.
The core of this project is powered by the ZevihaNut/3.0 neural network architecture, part of the aertsimon90/Zevihanthosa framework.
- Gender-Aware Biometrics: The system utilizes three distinct neural networks: a Male Model, a Female Model, and a Gender Classifier.
- Multi-directional Prediction: The model doesn't just predict from age. It understands the correlation between all variables. If you provide only age, it predicts height and weight; if you provide height and weight, it can estimate age and gender.
- Powered by Zevihanthosa: Leverages a custom-built
Brainstructure instead of traditional heavy frameworks, allowing for lightweight yet powerful regression. - Hybrid Interface:
-
GUI (Graphical User Interface): Built with Tkinter for interactive sliding scales, real-time visualization, and gender toggles.
-
CLI (Command Line Interface): Robust terminal support for automated training, data entry, and quick predictions.
-
Visual Regression Analysis: Integrated Matplotlib "Review" feature to visualize how well the AI line fits the training data points for both genders simultaneously.
The system normalizes all inputs to a range between 0.0 and 1.0 before processing. The architecture uses a specialized "Multi-Brain" logic:
-
The Gender Brain: Classifies the input as Male (0.0) or Female (1.0).
-
The Biometric Brains: Two separate
[[3, 12], [12, 6, 3]]networks trained specifically on male and female growth curves.
- Autoencoder Logic: During training, the system trains on every permutation of input (Age only, Height+Weight, etc.) to allow the model to "fill in the blanks" regardless of which parameter is missing.
Simply run the script without arguments to launch the dashboard:
- Predict: Adjust sliders. If you set the Gender to "Blank (Predict)", the AI will first guess the gender and then use the corresponding model to predict other metrics.
- Train: Open the training window to run additional epochs.
- Review: Generates a statistical plot comparing the Male (Blue) and Female (Red) prediction curves against actual data points to calculate the Overall Accuracy.
Use the terminal for quick operations:
- Create a New Model:
python main.py new --learning=0.01 --maxage=120 --pretrainepoch=50
- Add Custom Data (Age, Height, Weight, Gender[0=M, 1=F]):
python main.py add 25 180 75 0
- Quick Prediction (Gender: 0, 1, or 0.5 for Auto-Detect):
python main.py guess 20 175 0 0.5
The "Review" module calculates the Mean Absolute Percentage Error (MAPE). By comparing the AI's guesses for every point in the training set against the actual values, it generates an Accuracy Score:
This provides full transparency into the model's reliability across different age groups.
Distributed under the MIT License. See LICENSE for more information.
Predicting the human form, one epoch at a time. 🚀