ISOFRAMES
GitHub

SENTINEL

Real-time contextual safety intelligence platform for vehicles

Source on GitHub All systems
Status
Research
Version
No tagged release
Platform
GPU workstation
Source
github.com/Sherin-SEF-AI/Sentinel-Autonomy-Phase1
License
Not specified
01 · Problem

The README frames the goal as reducing accidents by understanding both the environment and the driver's state together, correlating external threats with driver awareness.

02 · Architecture

How it is put together.

SENTINEL is a contextual safety platform for vehicles that combines a multi-camera 360° Bird's Eye View with a driver monitoring system. A contextual intelligence engine correlates environmental threats with driver awareness and drives alerts adapted to the driver's cognitive load. The repository contains Python modules for camera capture and calibration, perception (BEV, segmentation, detection), driver monitoring, risk assessment, alerts, recording, CAN bus telemetry, HD maps, driver profiling and cloud sync, with a PyQt6 desktop GUI and a FastAPI/WebSocket dashboard.

Camera managementsrc/camera/Perceptionsrc/perception/Driver monitoringsrc/dms/Contextual intelligencesrc/intelligence/Alert systemsrc/alerts/CAN bussrc/can/GUI and dashboardsrc/gui/Camera managementsrc/camera/Perceptionsrc/perception/Driver monitoringsrc/dms/Contextual intelligencesrc/intelligence/Alert systemsrc/alerts/CAN bussrc/can/GUI and dashboardsrc/gui/
FIG. 01Modules and data flows, from the repository layout and documentation.
  1. 1

    Camera management

    Capture, synchronisation and calibration

    src/camera/
  2. 2

    Perception

    BEV, segmentation and detection pipeline

    src/perception/
  3. 3

    Driver monitoring

    Face, gaze, pose, drowsiness, distraction, readiness

    src/dms/
  4. 4

    Contextual intelligence

    Scene graph, risk, TTC, trajectory prediction

    src/intelligence/
  5. 5

    Alert system

    Alert generation, suppression and dispatch

    src/alerts/
  6. 6

    CAN bus

    DBC parser, telemetry reader, command sender

    src/can/
  7. 7

    GUI and dashboard

    PyQt6 desktop GUI and FastAPI/WebSocket web dashboard

    src/gui/
03 · Capabilities

Specification.

Cameras
3 USB 3.0 cameras (interior, front-left, front-right) with chessboard calibration script
Perception
BEV generation, semantic segmentation, multi-view YOLOv8 detection with Kalman tracking
Driver monitoring
Attention, drowsiness and distraction tracking using MediaPipe face landmarks
Intelligence
Scene graph, contextual risk assessment, time-to-collision and trajectory prediction
Vehicle bus
CAN interface with DBC parsing, telemetry reading and command sending
Interfaces
PyQt6 desktop GUI; FastAPI + WebSocket web dashboard with playback
Validation
pytest suites for integration, performance, reliability and accuracy (scripts/run_validation_suite.py)
Minimum hardware
NVIDIA GPU (compute capability >= 7.0, 8 GB+ VRAM), 8-core CPU, 16 GB RAM
Runs on
README requires an NVIDIA GPU with 8 GB+ VRAM; requirements-gpu.txt/requirements.txt use onnxruntime-gpu and torch
Stack
  • Python
  • PyTorch
  • ONNX Runtime (GPU)
  • Ultralytics YOLOv8
  • MediaPipe
  • OpenCV
  • filterpy
  • PyQt6
  • PyQtGraph
  • FastAPI
  • Uvicorn
04 · Related

More systems.

Perception and ADAS · LiveLyne ADAS

Real-time, on-device driver assistance for any Android phone.

Perception and ADAS · LiveDashLab

Train object detectors and visual models on dashcam footage without labelling any of it

Next systemAutonomous Vehicle Perception System

Multi-camera ADAS platform for autonomous vehicle development