Skip to main content

Edge Deployment

Hardware Requirements

ComponentMinimumRecommended
Edge DeviceJetson NanoJetson Xavier NX
RAM4GB8GB
Storage32GB64GB
Camera Resolution720p1080p
Frame Rate15 FPS30 FPS

Supported Cameras

Camera TypeCompatibilityNotes
IP CamerasRTSP streamsMost common
USB CamerasDirect connectEdge device local
Existing CCTVVia encoderRetrofit existing
PTZ CamerasSupportedPosition presets

Edge Processing

All vision AI runs on-premise:

AdvantageDescription
Low LatencyUnder 100ms processing
PrivacyVideo never leaves premises
Offline CapableWorks without internet
BandwidthNo video upload required

Cloud Sync

Only metadata syncs to cloud:

Data TypeSyncedFrequency
Detection EventsYesReal-time
Aggregated MetricsYesEvery 5 min
Model UpdatesYesDaily (download)
Video FootageNoNever
Still ImagesOptionalOn-demand

Best Practices

  1. Camera placement - Mount cameras to avoid glare and obstructions
  2. Lighting - Ensure consistent lighting for better detection
  3. Model training - Train on YOUR menu items for best accuracy
  4. Regular calibration - Recalibrate after camera moves
  5. Privacy notices - Post appropriate signage for customers
  6. Edge maintenance - Keep edge devices updated and ventilated

Troubleshooting

Low Detection Accuracy

Symptoms: Missed items, false positives

Solutions:

  1. Check camera focus and cleanliness
  2. Verify lighting conditions
  3. Retrain model on recent images
  4. Adjust confidence thresholds

High Latency

Symptoms: Slow detection, lag in updates

Solutions:

  1. Check edge device CPU/GPU usage
  2. Reduce camera resolution if needed
  3. Verify network connection to edge
  4. Clear edge device cache