🎯 OKITO Features

Complete overview of AI video surveillance platform features: object detection, pose analysis, workplace monitoring, notifications, and archive.

🔍 Object Detection

🎯

Detection Model

Modern models for accurate real-time object detection. Support for 80+ COCO classes.

  • person — human
  • car, truck, bus — vehicles
  • cell phone — phone
  • smoke, cigarette — smoking
📐

Sliced Detection

Image slicing technology for detecting small objects on high-resolution cameras.

  • Slice: 128px
  • Overlap: 0.6
  • Interval: every 2nd frame
  • Small Area Ratio: 0.08
🎚️

Sensitivity

Three sensitivity levels for different use cases.

  • High — minimum false positives
  • Medium — balanced mode
  • Low — maximum sensitivity
💡 The detection model can be adapted to the customer's needs. Training on custom datasets, adding custom classes, optimization for specific scenarios.

🧍 Pose Analysis NEW

🏃
Action Detection
Human pose detection using 17 COCO keypoints.
  • ✓ sitting
  • ✓ standing
  • ✓ walking
  • ✓ arm_bent
  • ✓ object_not_moving
📱
Static Pose Phone
Indirect phone detection by pose. Model: Pose Detection Model.
  • ✓ Hold: 3-6 sec
  • ✓ Motion max: 0.01-0.04
  • ✓ Elbow angle: 100-130°
  • ✓ Cooldown: 20-40 sec

📱 Phone Detection

Advanced phone detection logic with class separation by position and context.

Class Description Threshold Status
on_phone Phone at ear 65% ✓
phone_at_ear Phone at ear (detailed) 65% ✓
phone_in_hand Phone in hand 60% ✓
possible_phone Possible phone 55% optional
smoking Smoking (if not phone) 70% ✓

📍 Detection Zones

🗺️
Polygon Zones
Limit detection area. Events are recorded only within zones.
  • ✓ DRAW ZONE — start drawing
  • ✓ CLOSE ZONE — finish and save
  • ✓ REMOVE LAST — remove last
  • ✓ RESET — clear all
💼
Workplace Zones
Separate zones for workplace monitoring with independent incidents.
  • ✓ Absence alert: 30 sec
  • ✓ Presence confirm: 5 sec
  • ✓ Require initial presence
  • ✓ Per-zone incidents

📲 Telegram Notifications

🔔

Instant Alerts

Notifications at the moment of event with photos and context.

  • Cooldown: 60 sec between events
  • Batch: grouping within 5 sec
  • Screenshots: before/after event
  • Confidence in percent
📸

Screenshots

Automatic camera screenshots with each notification.

  • Before: 2 frames
  • After: 2 frames
  • Interval: 3 sec
  • Size: 640×480

Notification Format

🚨 #smoking 📍 Camera: Office Entrance ⏰ 14:25:33 🎯 Confidence: 87%

📼 Archive & Timeline

📹

Auto Recording

Automatic recording when detections are present.

  • FPS: 20
  • Quality: MJPEG 80
  • Segments: 2 sec
  • Stop: +5 sec after event
📊

Timeline UI

NVR-like interface for archive navigation.

  • Segments by duration
  • Synchronized playhead
  • PREV/NEXT navigation
  • 48H / 7 DAYS modes
  • Mouse wheel zoom
⚡

Events Panel

All detections for selected period with filtering.

  • Search by date/time/type
  • Confidence in percent
  • Filter by event type
  • Click → jump + autoplay

Event Colors on Timeline

smoking, fall
on_phone, phone_at_ear
workplace_absence
sitting
standing