shukkkur / workout-tracker

Workout Tracking using Pose-Estimation

Workout Tracking 🏋️

Count the number of repetitions for 3 exercises: bench press, overhead press and dumbbells curl

How it works

1. Using mediapipe process video, detect body landmarks and write it into the csv (data_engineering.py).
2. Train RandomForestClassifier on created csv with four classes: 'bench', 'curl', 'overhead' and 'other'(ml_model.py).
3. Process a new video, classify the detected pose using the model (rf_classifier.pkl) and based on the classification calculate the angle of elbow and shoulder (calc_angle.py).
4. If the calculated angle exceed a certain threshhold, increase counter by 1.

How to run

1. Clone this repository

git clone https://github.com/ArtLabss/workout-tracking.git

2. Install the requirements using pip

pip install -r requirements.txt

3. Run the following command in the command line

python prediction.py --input_video_path=Input/bench_test.mp4 --output_video_path=Output/video_output.mp4 --draw_pose=0 --info=0

• --input_video_path - path to the input video
• --output_video_path - path to the output video, if not specified the video will be created in the same directory as input video
• --draw_pose - if set to 1 the detected poses will be drawn, default is 0
• --info - if 1 the probabilities of pose classification ([[prob_bench, prob_curl, prob_overhead, prob_other]]), elbow and shoulder angles, as well as which hands is used for calculating the angle will be shown, default is 0



If you stumble upon an error or have any questions feel free to open a new Issue