Discover how we helped businesses transform their operations with AI automation. Real results, measurable impact, and proven ROI across multiple industries.
YOLOv8-based safety monitoring running on edge devices (NVIDIA Jetson) inside the steel plant — 24/7 detection of helmet and high-visibility vest compliance with real-time alerts to the safety team.
Manual safety patrols cannot cover the whole plant 24/7. Off-the-shelf PPE-detection products struggled in heavy industry conditions: poor lighting, dust, partial occlusions and glare from molten steel.
Streaming raw camera feeds to a central server was not an option due to bandwidth and data-residency requirements.
A fine-tuned YOLOv8 detector trained on plant-specific footage, deployed to NVIDIA Jetson devices co-located with each camera.
The model runs locally, emits structured detection events (timestamp, camera ID, violation type, confidence), and only metadata leaves the edge device. A central dashboard aggregates events and routes alerts to the safety team.
Per-camera inference on Jetson Orin runs the fine-tuned YOLOv8 model at real-time frame rates.
Detection events are pushed to a central event bus (MQTT), filtered through a debouncing rule engine to suppress flicker, then surfaced in a Grafana-style operations dashboard.
No raw video leaves the plant network; only event metadata and small verification snapshots are retained.
