Running-Form-Correction

Utilizes Pose Estimation to offer sprinters cues based on an image of their running form.

How to Run

Dependencies

You will need the dependencies listed below: Note: it is encouraged that you utilize a venv through either pip or anaconda

• python3
• tensorflow 1.3
• opencv3
• protobuf
• python3-tk

Install

$ git clone https://github.com/dfrdeleon/Running-Form-Correction
$ cd Running-Form-Correction
$ pip3 install -r requirements.txt

Pose Estimation Demo

To see an example of the pose estimation overlayed on top of the original image, run the code below. Note: Inference generation works best if only one person and their entire body is in frame.

You can set model equal to one of the networks listed below:

• cmu
• dsconv
• mobilenet
  • mobilenet_fast
  • mobilenet_accurate

Make sure to set the imgpath to that of the input frame on your machine.

$ python3 inference.py --model=cmu --imgpath=...

Form Correction

Similar to the Pose Estimation Demo, set the intended model and image path:

$ python3 form_correction.py --model=cmu --imgpath=...