Inspiration
Falls are a serious problem, especially for elderly people living alone or patients in hospitals. If someone falls and no one notices right away, it can lead to dangerous situations.
We wanted to explore how computer vision could help solve this problem. Our idea was to create a simple system that can monitor a person using a camera and automatically detect if a fall happens. If it does, the system sends an alert so someone can respond quickly.
What it does
Our project is a real time fall detection system powered by computer vision.
The system uses a webcam to monitor a person and track their body posture using pose detection. By analyzing how the body moves and the angle of the person's posture, the program can detect when someone suddenly falls or ends up lying on the ground.
When a fall is detected, the system automatically triggers an emergency alert using the Twilio API. This could notify caregivers, family members, or emergency contacts.
How we built it
We built the system using Python with OpenCV and MediaPipe.
MediaPipe Pose allows us to detect and track key body points such as shoulders, hips, and joints. Using those points, we calculate body orientation and movement patterns to determine if a fall might have occurred.
OpenCV processes the video stream from the webcam in real time, while Twilio handles the alert system that gets triggered when a fall is detected.
Challenges we ran into
One challenge was figuring out how to distinguish between normal movements (like sitting down or bending) and an actual fall.
We experimented with different body angle and movement thresholds to reduce false alarms. Another challenge was getting the real-time camera processing to run smoothly while still analyzing pose data.
Integrating the alert system with the detection logic also required some debugging.
Accomplishments that we're proud of
We're proud that we were able to build a working real-time prototype during the hackathon.
The system successfully detects falls using pose estimation and can trigger an emergency alert automatically. Bringing together computer vision and an alert system in such a short amount of time was a great achievement for us.
What we learned
Through this project we learned a lot about:
- Real-time pose detection using MediaPipe
- Working with OpenCV for computer vision
- Detecting events based on body movement data
- Integrating APIs like Twilio into our applications
- Building a full prototype under a tight time limit
What's next
If we continue developing this project, we would like to:
- Improve the accuracy using trained machine learning models
- Support detecting multiple people at once
- Send SMS or mobile notifications when a fall is detected
- Deploy the system on edge devices like Raspberry Pi
- Explore use cases in hospitals, assisted living facilities, and smart homes
Log in or sign up for Devpost to join the conversation.