For physical therapy patients, doing your home exercise program is a crucial part of therapy and recovery. These exercises improve the body and allow patients to remain pain-free without having to pay for costly repeat visits. However, doing these exercises incorrectly can hinder progress and put you back in the doctor’s office.

What it does

PocketPT uses deep learning technologies to detect and correct patient's form in a broad range of Physical Therapy exercises.

How we built it

We used the NVIDIA Jetson-Nano computer and a Logitech webcam to build a deep learning model. We trained the model on over 100 images in order to detect the accuracy of Physical Therapy postures.

Challenges we ran into

Since our group was using new technology, we struggled at first with setting up the hardware and figuring out how to train the deep learning model.

Accomplishments that we're proud of

We are proud that we created a working deep learning model despite no prior experience with hardware hacking or machine learning.

What we learned

We learned the principles of deep learning, hardware, and IoT. We learned how to use the NVIDIA Jetson Nano computer for use in various disciplines.

What's next for PocketPT

In the future, we want to expand to include more Physical Therapy postures. We also want to implement our product for use on Apple Watch and FitBit, which would allow a more seamless workout experience for users.

Built With

Share this project: