Helmet & vest, every person
Detects hard hats and hi-vis vests on each tracked person and flags anyone missing gear the moment they enter frame.
- Helmet detection
- Hi-vis vest detection
- Per-person compliance state
Safety & Compliance · Computer Vision
Safety monitoring that runs on-prem.
Boundary surveillance, PPE compliance and intrusion detection from your existing cameras — real-time video analytics that never leave the building. No cloud, no servers, no internet required.
0
Detection confidence threshold on every person and PPE call.
Zero cloud
Nothing is uploaded. All inference and storage stay on your hardware.
Runs offline
No servers, no internet. Deploys as a desktop application on-site.
What it watches
Three jobs, running on every frame.
Open Vision PPE watches your cameras for the specific things that get someone hurt — missing gear, people in places they shouldn't be, and headcount through a boundary.
Intrusion into no-go areas
Draw a boundary once. Any person crossing into a restricted zone raises a violation and is logged with the frame and track.
- Boundary surveillance
- Zone-violation detection
- Instant breach flags
Headcount across a line
ByteTrack follows each person across frames and counts direction of travel — a live, auditable tally of who went in and out.
- Multi-person tracking
- Direction-aware counting
- Live bounding boxes
How it runs
Detect → Track → Report.
A single pipeline turns raw camera frames into compliance events, tracked identities, and a report in your inbox — all on the same machine.
STEP 01
Detect
YOLOv8n runs person detection at 320×320 with a 50% confidence threshold, then evaluates PPE and restricted-zone state on each detection.
STEP 02
Track
ByteTrack assigns a stable ID to every person using IoU and a 30-frame history, while InsightFace 512-dim embeddings support face recognition where needed.
STEP 03
Report
Events and counts land in a local SQLite database. Daily PDF and Excel safety reports are generated and delivered over SMTP — no dashboard login required.
No cloud · No servers · No internet
Your footage never leaves the building.
Open Vision PPE is a desktop application. Inference, tracking and storage all run locally, so there is no upstream to secure, no bandwidth bill, and nothing to breach off-site.
- All data in SQLite, on-siteEvents, tracks and counts persist in a single local database file.
- Reports auto-emailed over SMTPDaily PDF and Excel summaries reach the safety team automatically.
- Runs on existing camerasWorks from the RTSP feeds you already have — no new hardware to certify.
DATA FLOW · ON-PREM
LOCAL ONLY
Existing cameraRTSP video feed
On-site desktop appYOLOv8 · ByteTrack · InsightFace
Local SQLite + reportsPDF / Excel, auto-emailed
Product tour
What the safety team actually sees.
A live monitor for the floor, and a full trail behind every violation — from the detection box to the auto-generated report.
01 · LIVE MONITOR
Every camera on one wall
The safety monitor shows live feeds side by side, with PPE state and restricted-zone overlays drawn on each person in real time.
- Live PPE and zone detection
- Per-camera IN / OUT counts
- Violations surfaced as they happen
02 · VIOLATION TRAIL
Every flag, fully traceable
Open a violation to see the detection boxes, the track filmstrip that identifies the person across frames, and the auto-report trail that gets emailed out.
- Detection boxes on the flagged frame
- Track filmstrip across 30 frames
- Auto-report trail to the safety team
Under the hood
Open vision models, wired for the floor.
Built in Python on proven open-source detection and tracking — no proprietary black box, no vendor lock-in.
Python
YOLOv8n 320×320
ByteTrack IoU · 30f
InsightFace 512-dim
SQLite
SMTP PDF · Excel
Deployed for an industrial safety client
Put every camera to work on safety.
See Open Vision PPE running on feeds like yours — PPE compliance, restricted zones and IN/OUT counting, entirely on-prem.