Exercise Counting Classification ML Model

A machine learning model capable of classifying 5 calisthenics exercises namely; Push-ups, Sit-ups, Squats, Jumping-jacks, and Planking using Mediapipe Pose.

This repository contains Scripts and GUI Applications that can be used to:

• Record a dataset using GUI App
• Script for Training Machine Learning Model
• Test the performance of the Model using Mediapipe Pose and opening a sample video
• Measure the angle of 3 points in 3 Dimensions using the X, Y, and Z values from Mediapipe Pose Landmarks

Dataset

The Dataset used in training the Model contains 9 labels:

Labels	Exercises
1	Situps Down
2	Situps Up
3	Pusups Down
4	Pushups Up
5	Plank
6	Squat Up
7	Squat Down
8	Jumping Jack Up
9	Jumping Jack Down

Model Performance

The score of the latest ML Model (exercisev3.pkl):

The Confusion Matrix produced through testing the prediction of the ML Model:

How to use the repository

1. Create a Python virtual environment
2. Activate the Python environment
3. Install the dependencies in requirements.txt

   pip install requirements.txt

4. Create your own Dataset using "datasetRecorder_GUI.py" GUI App
5. Train your model using the dataset or CSV file using "train_model.py"
6. Test your model using "exercise_test_counter.py"