ONA Control Center
Real-time spatial awareness & configuration
Camera Not Active
Click start to begin
Activity
System Status
Camera
Test Reports
Recording will start with detection
Settings
Recognition
Minimum confidence to recognize a face. Higher = fewer false positives but may miss some people. Lower = more detections but less accurate.
Controls initial face detection sensitivity. Higher = only obvious faces detected. Lower = detects distant/partial faces but more false positives.
Detection
RT-DETR: Higher accuracy, more detections (~200ms). YOLO: Faster inference (~40ms), lighter model.
Detects hazards, obstacles, and objects using YOLO. Enable for spatial awareness & navigation. Disable for face-only mode (faster).
Estimates distance to objects/people. Enable for distance-based filtering. Disable to detect at all distances (faster).
Hazards block face detection. Enable for safety-first (announces obstacles before people). Disable to run both in parallel.
Visualizes detected faces and objects. Enable to see what the AI sees. Disable for a cleaner view.
Uses advanced AI to humanize announcements. Enable for natural, conversational speech. Disable for faster, robotic alerts.
Brief and direct.
Performance
Process every Nth frame for faces. Higher = much faster (3x-10x) but may miss brief appearances. Lower (1) = every frame, slower but most accurate.
Downscale frames before detection. Lower = faster (25% = 16x speedup) but misses small/distant faces. 100% = full quality, slower.
Remember face recognition results. Higher = faster for static scenes (reuses results). 0s = always fresh, recalculate every time. Best: 1-2s.
Announcements
Minimum time between face announcements. Higher = less spam but may not update if person moves. Lower = frequent updates but can be annoying.
Minimum time between obstacle warnings. Higher = cleaner audio but may delay critical updates. Lower = safety-first, more frequent warnings.
Announces direction to avoid hazards (e.g., "Move left"). Enable for navigation help. Disable for location-only announcements.
🌍 Language
Changes voice announcements to selected language. Uses Piper TTS with native voices. Restart detection for changes to take effect.