Track people fast
DETR keeps the person boxes responsive while the larger vision model works on a slower cadence.
Pair real-time person boxes from DETR with a selectable vision model that can check for either hardhats or baseball caps.
Construction safety analytics
Start the camera to assess hardhat compliance with Workers AI.
DETR tracks people quickly; the selected vision model checks hardhats every few seconds.This view combines a fast detector with a slower vision-language model so the UI stays responsive while PPE checks complete.
DETR keeps the person boxes responsive while the larger vision model works on a slower cadence.
Choose Llama Vision, LLaVA, or Gemma 4 to reason about hardhats or baseball caps from the sampled frame.
The browser logs transitions when the frame changes from compliant to missing the selected headwear or back again.