Inspiration
Every year 360000 die by drowning. One of our friend almost died from drowning. This idea is inspired from him.
What it does
- We have build a monitoring system that keep tracks of people who are swimming.
- We built a machine learning model that identifies swimming, drowning and other activity.
- This system is also extended to monitor people who are suffering from epilepsy.
- The model is further trained on people who suffer from seizures and tries to alert emergency helpline.
How I built it
- We leveraged the accelerometer and gyroscope sensors attached to arduino board which provides us with yaw, pitch and roll data. This data is fed to a machine learning algorithm with an activity label. The model is trained on data thus collected.
- We created a node.js server that runs the machine learning model and listens to further input from the sensors. Based on which it predicts the activity of the person.
- The monitoring is done through a webapp that listens to the prediction data sent by the server. (You will see the magic)
Challenges I ran into
- Connecting the sensors to arduino board and sending the data via bluetooth module was a bit tedious.
- The data sent was not a continuous chunk but was data with bits and pieces.
- Creating a system to collect the data to train the model took some time.
- Understanding the data and deciding on the right machine learning model
Accomplishments that I'm proud of
- The system actually works with model accuracy around 88%.
- We managed to collect sufficient data to train the model.
- We also managed to predict with real time data.
What I learned
- Bluetooth communication with server.
- Different data sent by accelerometer and gyroscope sensors
- Creating a system that collects, train and predicts real time data.
What's next for SafeSwim
- Collect data from different people.
- Visualise their movement in a way so as to help them improve their swimming skills.
- Collect more data from epilepsy patients and connect to healthcare agencies for better service.
Log in or sign up for Devpost to join the conversation.