Why Greek Motorcycle Drivers Are Not Wearing Helmets?
summary
Cameras on balconies and public live streams feed a Node.js pipeline that extracts frames with ffmpeg and runs an on‑device image model trained in Google Colab using the Yellow (Ultralytics) library, exported to TensorFlow/TF‑Lite for live inference. The system runs six months of continuous data, counts about 3,000 motorbikes per day, blurs faces and plates, detects helmet/no‑helmet per rider, and finds a plausible weather‑related correlation rather than enforcement as the main factor.
AI-written summary · about
related talks
47mUnlocking Data-Driven Apps with TensorFlow
48mFrom "Hello World" to "Hello Smart City"!
19mAI Trends in Intelligent Transport Systems
21mAI Isn't Just Chat
28mBuilding Motion Controlled Websites in JavaScript
20mMask || No Mask? Detect Masks with ml5.js
41mTensorFlow.js 101: ML in the Browser and Beyond
29mLife on the Edge