A real-time AI-powered fall detection system that monitors people through webcams and detects potential falls and accidents using computer vision and pose estimation. Built with Python, Flask, javascript and MediaPipe.
It watches people through a webcam and uses AI to track their movements. When it detects someone falling - whether it's a rapid drop, sudden impact, or someone ending up lying down - it immediately sends alerts to family or close friends, or in the case of a serious emergency, 911. It is in essence, like having a smart security camera that can tell when something's wrong.
This system was built specifically with elderly care and disability support in mind. Falls are the leading cause of injury and hospitalization for older adults, and many people with mobility challenges live alone or spend significant time unsupervised. Traditional monitoring systems often require wearable devices that people forget to charge or find uncomfortable.
For Elderly Care:
- Monitor aging parents or grandparents who live independently
- Detect falls in assisted living facilities and nursing homes
- Provide peace of mind for family members who can't be present 24/7
- Reduce response time to medical emergencies
- Track daily movement patterns to identify health changes
For People with Disabilities:
- Monitor individuals with mobility impairments who may be prone to falls
- Support people with balance disorders or seizure conditions
- Assist caregivers in monitoring multiple people simultaneously
- Provide independence while maintaining safety oversight
- Detect unusual movement patterns that might indicate medical issues
Other Applications:
- Home healthcare monitoring
- Rehabilitation progress tracking
- Post-surgery recovery monitoring
- Child safety in high-risk areas
- Workplace safety in hazardous environments
The beauty of this system is that it's completely non-intrusive - no one needs to wear anything or remember to activate it. It just works in the background, providing a safety net that can literally save lives.
The fall detection uses several different methods to catch falls. It tracks how fast someone's moving down, detects sudden impacts, and notices when someone's body goes from standing upright to horizontal. The system analyzes 33 different body points in real-time to figure out what's happening.
You get a live video feed with the AI's analysis overlaid on top. When a fall is detected, the screen flashes red and you get clear alerts. The system automatically clears the alert after a few seconds once it thinks the person has recovered.
You'll need Python 3.12 or higher and a webcam. The project uses uv for fast and reliable package management.
First, install the dependencies:
uv syncThen run the app:
uv run app.pyAccess it at http://localhost:5001
Click "Start Camera" in the web interface to begin monitoring. Position yourself in front of the camera - the system will automatically start tracking your movements. If it detects a fall, you'll see immediate alerts and the video feed will get a red border.
The interface shows your current stats, including whether a fall was detected and when. You can start and stop monitoring anytime, and clear fall alerts when needed.
You can tweak the detection sensitivity and camera settings in the app.py file. The main settings control how confident the AI needs to be before it triggers alerts, alongside basic camera settings like brightness and contrast.