Visenious

Visual Super Intelligence. Cameras that report, not just record.

Dynamic camera security, imagery analysis, face recognition and document understanding on your site. Compact events with evidence. Video never leaves.

Email usThe four capabilities

Four capabilities. One evidence contract.

Every event: what, where, when, which model, how confident, and a hash of the evidence.

  • VISUAL / 1Alerts

    Dynamic camera security

    Zones, dwell, intrusion, alert.

    Live detection and tracking on your streams, with zone and dwell rules.

    • Alerts with evidence, rule and time.
    • Degraded coverage flagged per camera.
    • Routed to the operator's channel.
    • Tracking never establishes identity. Attendance uses trusted badge records by default.
  • VISUAL / 2Reports

    Imagery analysis

    From pixels to findings.

    Counting, classification, condition and change on stills and footage.

    • Specialist models per domain.
    • Batch reports for inspection, inventory and compliance.
    • Findings feed VSI and VPSA SI as evidence.
  • VISUAL / 3Per-site assessment

    Face recognition

    Identity, under a defined purpose.

    Enrollment-based identification for access, attendance and security.

    • Only with a purpose, a legal assessment and retention rules.
    • Evaluated across lighting, angle and occlusion.
    • Local templates, local matching.
  • VISUAL / 4With VSI

    On-demand understanding

    Photos and documents become validated facts.

    Submitted photos, screenshots, diagrams and scanned documents, read with OCR and multimodal reasoning, checked against the ERP record.

    • Page and region references kept for every extracted fact.
    • A damaged carton at receiving becomes an inspection case linked to its order.
    • A detection alone never changes the ledger. Qualified evidence and authorised review do.

Events, not frames.

Perception at the edge. Reasoning in the core.

  • 2.1 Edge

    Inference stays on site.

    Decode, detect, track and debounce on your hardware.

  • 2.2 Evidence

    Every event carries its proof.

    Camera, time, model, confidence and a hash of the clip.

  • 2.3 Delivery

    An outage never drops an event.

    A local outbox retries until the core accepts.

  • 2.4 Alerts

    To the right person.

    Evidence attached. Acknowledgment recorded.

  • 2.5 Reports

    Numbers with denominators.

    Precision, recall and false alerts per camera-hour.

  • 2.6 Governance

    Alerts first.

    Physical responses need their own authorization.

A breach at 02:14. Evidence attached.

  • Confidence is detector output, not identity.
  • Disposition is recorded on the same event id.
visual · site east gate · cam 07live
  • $ visual watch --camera 07 --zone perimeter-east
  • ⟶ detect person track #4412 dwell 7.2 s conf 0.88 model yolo-v11.3
  • ⟶ evidence clip 02:14:03–02:14:11 sha256 e66a…90b
  • ⟶ deliver outbox → ingress accepted event #88 120
  • genius: no scheduled work on the east perimeter. Proposal: alert security lead, hold gate 2.
  • ⟶ govern alert:send → pass · gate:hold → reviewer required
  • ✓ alert 02:14:14 · acknowledged 02:15:02 · patrol dispatched · report #88 120

Qualified per site.

Cameras, rates, retention and error budgets are agreed and tested per deployment.

Streams
Approved RTSP sources.
Models
Pinned, evaluated on your footage.
Compute
Shared GPU, dedicated GPU or edge device.
Evidence
Local object store. Your retention.
Face recognition
Optional module with its own purpose and legal assessment.
Runs on
Genius. Nothing leaves the site.

Also from Visenious

Show us your cameras.

One email to list the streams and the rules.

Email us