Automotive Occupant Detection and Cabin Safety System
SEAT BELT Desk Application Guide
SEAT BELT Desk analyzes cabin video to count occupants, report seatbelt use for each visible person, and flag driver attention events. Operators can monitor a live browser camera or analyze a recorded journey and review the results afterward.DragonFlyTest transforms live camera feeds and recorded cabin videos into occupant insights, intelligent alerts, and visual reports that support safer driving and proactive risk monitoring.
Monitor Occupants and Seatbelt Compliance
Detect drivers and passengers inside the vehicle, count visible occupants, and identify individual seatbelt usage. Colour-coded indicators display Belt on and Belt off status for each detected occupant.
Detect Driver Fatigue and Sleep
Use AI-powered facial landmark analysis to identify eye closure, yawning, and head tilting. Generate timely warnings when potential fatigue-related behaviours are detected.
Identify Driver Distraction and Unsafe Posture
Analyse facial orientation and body posture to identify potential driver distraction and slouching. Highlight attention-related events to support timely safety reviews.
Our Approach
DragonFlyTest Our Automotive Cabin Safety approach combines AI-powered occupant detection, seatbelt monitoring, and driver-fatigue analysis to identify potential risks in real time. We transform live and recorded cabin footage into intelligent alerts and actionable safety insights.
01.Cabin Monitoring & Mode Selection
Configure the application to monitor occupants, seatbelt compliance, and driver attention.
Select Seatbelt, Sleep, or Merge monitoring mode.
Enable real-time occupant detection and cabin monitoring.
Configure seatbelt and driver-fatigue analysis.
Combine safety checks based on the selected monitoring mode.
02.Occupant & Seatbelt Detection
Identify cabin occupants and evaluate individual seatbelt usage.
Detect drivers and passengers using YOLOv8 person detection.
Assign occupant roles such as Driver and Front Passenger.
Detect seatbelt straps using a dedicated YOLOv8 model.
Display Belt on in green and Belt off in red for each occupant.
03.Driver Attention & Fatigue Analysis
Monitor visible driver behaviours to identify potential fatigue and distraction indicators.
Detect eye closure using facial landmarks.
Identify yawning through mouth-opening measurements.
Calculate head tilt using facial and pose landmarks.
Analyse driver distraction and posture conditions.
04.Real-Time Alerts & Safety Review
Generate safety warnings and visual records to help operators review detected events.
Trigger alerts for seatbelt violations and driver-attention events.
Activate audible sirens for supported fatigue warnings.
Capture annotated snapshots of detected safety events.
Review alert history, snapshots, and event details through the dashboard.
Occupant Detection Features
- Detects people inside the cabin and ignores non-person detections in the final cabin pipeline.
- Filters very small background person boxes to reduce false occupant counts.
- Tracks person boxes across consecutive frames using box overlap.
- Requires passenger detections to persist before adding them to the cabin layout.
- Keeps the driver visible immediately while passenger detections are being confirmed.
- Counts the number of visible cabin occupants.
- Assigns Driver, Front passenger, Rear left, Rear right, and Rear center roles.
- Treats the rightmost visible occupant as the driver in the current camera layout.
Seatbelt Detection Features
- Uses a dedicated seatbelt model stored in models/yolov8n_seatbelt.pt.
- Checks that the loaded model is a strap detector and replaces invalid weights when possible.
- Recognizes seatbelt, strap, buckle, worn-belt, and unworn-belt class-name patterns.
- Matches each detected strap with the corresponding occupant upper torso.
- Checks the strap center and torso overlap before accepting Belt on.
- Rejects extremely short or narrow detections that may be shirt folds or creases.
Driver Sleep Detection Features
- Uses face landmarks around both eyes to calculate the eye aspect ratio.
- Detects closed eyes when the eye ratio falls below the configured threshold.
- Uses recent driver-attention readings to reduce one-frame changes.
- Produces Driver Sleep Warning when the sleep condition is confirmed.
- Shows the current eye ratio on the live camera overlay.
- Draws visible eye landmark points for live diagnostic confirmation.
- Creates a warning event and annotated snapshot for a confirmed sleep warning.
- Activates the browser siren for a confirmed driver sleep warning.
Head Tilt Detection Features
- Uses the driver left-eye and right-eye pose landmarks to calculate the eye-line angle.
- Detects sideways head tilt when the absolute eye-line angle exceeds the configured limit.
- Produces Head Tilting Alert when the condition is confirmed.
- Displays the measured head-tilt angle on the live camera.
- Draws the eye line used for the tilt calculation.
- Creates a warning event, annotated snapshot, colored banner, and audible siren.
Yawn Detection Features
- Uses face landmarks around the lips to calculate a mouth aspect ratio.
- Detects a yawn when mouth opening exceeds the configured threshold.
- Produces Yawn Detected after the active attention reading is confirmed.
- Displays the current mouth ratio on the live camera.
- Draws mouth landmarks and guide lines for visual diagnostics.
- Creates a warning event, snapshot, warning banner, and audible siren.
Why DragonFlyTest
AI-Powered Automotive Cabin Safety Built Around Your Needs
DragonFlyTest combines artificial intelligence, computer vision, and intelligent video analysis to deliver automotive occupant detection and cabin safety solutions. From real-time occupant monitoring to seatbelt compliance and driver-fatigue detection, we focus on accurate visual insights, timely alerts, and simplified safety monitoring.
Intelligent Occupant & Seatbelt Detection
Detect drivers and passengers using YOLOv8 and monitor individual seatbelt usage with dedicated seatbelt detection, occupant tracking, and colour-coded status indicators.
Comprehensive Driver Safety Monitoring
Identify potential driver-fatigue indicators, including eye closure, yawning, head tilting, distraction, and improper posture using MediaPipe facial and pose analysis.
Flexible Monitoring & Video Analysis
Choose Seatbelt, Fatigue, or Merge mode to suit your monitoring requirements. Analyse live browser camera feeds or upload recorded cabin videos for safety assessments.
Real-Time Alerts & Visual Evidence
Receive critical seatbelt violation alerts and driver-attention warnings with audible sirens, annotated snapshots, and event details for easier safety review.
Centralised Dashboard & Safety Insights
Monitor occupant counts, seatbelt status, driver posture, and attention through an interactive dashboard. Review alerts, snapshots, and individual occupant details in one place.
Flexible Processing & Reliable Performance
Benefit from automatic GPU acceleration with CPU fallback, cached detection models, efficient frame processing, and configurable seat-zone settings for streamlined cabin monitoring.