Software that detects bad exercise techniques when analyzing the posture of the person

The system is able to analyze the technique of 4 weightlifting exercises: Bicep Curl, Triceps Pushdown, Shoulder Press and Front Raise.

This project combines machine learning and recent advances in pose estimation. Built with Python, Scikit-learn, SciPy and OpenPose.

In order to run this program, OpenOpose must be installed on the machine. You can download the required release here. After unzipping the "zip" file, navigate to the "models/" directory. You should see a windows batch file named "getModels". Click this file in order to download openpose pose estimation models. After completing these steps follow the steps below.

Steps to run the program:

• Download the repository.
• Move the "IndividualProject" folder in the directory where "openpose" folder exists. IndividualProject and openpose folders must be in the same directory for the software to run. See the image below:



• Navigate to the IndividualProject folder in the command prompt.
• Type "python Main.py --mode evaluation --exercise [bicep_curl OR shoulder_press OR front_raise OR triceps_pushdown] --video_path [path to the exercise video you want to analyze] --keypoints_folder [folder where the extracted keypoints should be stored (must be inside IndividualProject folder)]

• In the end you should get the output of the software showing the detailed analysis of each repetition performed in the video as shown below:

....