Kata | Devpost

Inspiration

Ask yourself: Do you feel safe walking alone on the road at 11 pm at night? We don't. That's why we built Kata. As more girls go to universities and colleges, and women join the workforce, their exposure to attacks of assault have been increasing day by day. At the same time, some women do not have the time or cannot afford going to self defense classes. Women have the right to their own safety, and Kata aims to connect them to this basic right. Bridging the gap between technology and safety, Kata teaches self defense to women, but with computer vision.

What it does

Kata extracts self defense poses from YouTube and TikTok tutorials. It then converts these poses into computer vision stick-figures, that you (the user) must match to perfect your self defense moves. We also developed a metric that shows the user a score and how they are performing. This makes learning self defense more engaging and gamifies martial arts with real-time feedback.

How we built it

Mediapipe's API paired with OpenCV to extract pose from user's webcam and self defense tutorial
Pickle to compile data
Python and flask for the backend of the application
HTML, CSS, JS and Bootstrap for the frontend of the application

Challenges we ran into

Python libraries and mediapipe would not get installed on the computer with pip install (we had to build a virtual environment on our terminal)
OpenCV was not allowed to access camera on Mac (followed long YouTube tutorials to change security preferences)
The pose extracted from the video would not display properly on the screen for the user to match (had to give it a scale)
Mediapipe slowed down the tutorial videos after extracting keypoints

What we learned

How to use Mediapipe and OpenCV for pose estimation
Various python libraries like numpy, pickle, imageio
Karate moves
How to use flask for backend

What's next for Kata

Authentication so user can login and sign up, and increase security as young girls would user Kata
Estimate 2 poses at the same time (attacker, defender)
Add a database of videos from YouTube and Tiktok
Add a yoga for mental health section where user can learn yoga
Build a chatbot that increases motivation to keep learning self defense

Built With

Submitted to

Hack with Us - TechNova 2022

Created by

I lead my team since I was the only hacker with previous hackathon experience. I worked on all of OpenCV and Mediapipe API for pose extraction and estimation. I learned new python libraries. It was also my first time learning backend, for which I learned flask.

Vidhi Ruparel
This was my first hackathon! I worked on the front-end and back-end development for our project, mainly on using the OpenCV and mediapipe for user pose extractions. It was also my first time working with flask and mediapipe.

ariishaat Ishaat
I built the website from scratch using flask and bootstrap. It was exciting to learn Flask and use it immediately. I also learned a lot about opencv and mediapipe.

Rujing Li
This was my first hackathon! I contributed to both the front-end and the back-end. It was my first time using MediaPipe and Flask. It was super cool to learn something completely new in 36 hours!

Kiran Mandhane