Fernando
A posture-correcting robot that taps you when you slouch
Overview
Fernando is a hardware robot designed to help you maintain proper posture while working at your desk. Using computer vision to detect your posture, Fernando physically corrects bad habits by gently tapping your hand when you slouch. Unlike apps that are easy to ignore, Fernando provides immediate physical feedback that's impossible to miss, helping you develop better posture habits over time.
The Problem
Poor posture during long hours at a desk leads to chronic back pain, neck strain, and other health issues. Most existing solutions rely on passive notifications that are easy to ignore, resulting in continued poor habits.
The Solution
Fernando provides immediate physical feedback when your posture deteriorates. The gentle tap from its mechanical arm creates an unmistakable reminder that helps reinforce good habits through consistent, timely correction.
Key Features
Real-time posture detection using OpenCV and MediaPipe to track 33 key body points, with 94% accuracy in identifying poor posture.
A robotic arm with force-calibrated tapping mechanism, designed for comfort while providing effective tactile feedback.
Next.js web application for tracking posture history, customizing sensitivity, and providing personalized improvement recommendations.
How It Works
Detection
Camera monitors your posture in real-time
Analysis
AI determines if your posture needs correction
Correction
Mechanical arm delivers a gentle tap
Learning
System adapts to your habits over time
Technical Details
Hardware Components
- Raspberry Pi 4 (4GB) for processing
- HD webcam for posture detection
- 3 high-torque servo motors for arm movement
- Custom 3D-printed PLA housing and arm
- Silicone padding for comfortable contact
- 5000mAh LiPo battery for 8-10 hours of operation
- USB-C charging port
Software Stack
- Python for core processing and control
- OpenCV and MediaPipe for pose estimation
- TensorFlow Lite for optimized edge inference
- Next.js and React for the web interface
- Vercel for application hosting
- WebSockets for real-time communication
- Custom microservices architecture
Results
Improved posture awareness
Measurable posture improvement
Preferred over app notifications
Rated the feedback as comfortable
Based on a 4-week study with 25 participants using Fernando in their daily work environment.
More Projects
View all Projects →
TikTok View Predictor
A machine learning-powered model that predicts the number of views a TikTok video will receive using linear regression.
