Sporty

Sporty is an AI model designed to assist during workout sessions by monitoring repetitions and movements using the Mediapipe library.

Environment and Requirements

Ensure you have Python 3.10.6 installed before proceeding.

1. Create a virtual environment using Python 3 (follow the installation instructions at virtualenvwrapper documentation):

   virtualenv <name_env>
   

2. Activate the virtual environment:

   source <name_env>/bin/activate
   

3. Install Python requirements:

   pip install -r requirements.txt
   

Important: The latest PyAV versions are incompatible with the pytorch version used. Info: pytorch/vision#7305

Data

Sporty utilizes the Workout Video dataset from Kaggle, which consists of 22 different categories of workouts.

Usage

Sporty offers two modes: with or without exercise prediction.

Run the main script:

python main.py

Without Exercise Prediction

In this mode, it's not necessary to have the model pre-trained beforehand. Currently active exercises include:

• Pull Up
• Squat

Exercise Prediction

To use this mode, where Sporty predicts exercises, it's necessary to train the model first. Follow these steps:

1. Preprocess Videos: Run preprocess_videos.py.
2. Split Data: Run split_data.py.
3. Train Model: Run train.py.

Next Steps

• Incorporate correction mechanisms for new exercises.
• Ensure predicted exercise names match the defined exercises.

Feel free to contribute and enhance Sporty's capabilities!