Preset: Deploy to Jetson {#default}

Run AI-accelerated person intrusion detection on your reComputer Industrial (Jetson Orin), with live video and configurable safety rules in a browser dashboard.

What you'll get:

• Live annotated video stream (bounding boxes + zones + lines) in the browser
• Configurable restricted zones, fence lines, and dwell-time rules
• Event log with snapshots when rules trigger
• HTTP API for stats / events / live MJPEG (/api/stats, /api/events, /api/stream)

Requirements: Jetson with JetPack 6.x · Docker + NVIDIA runtime · SSH access · Working RTSP camera URL

Step 1: Initialize Camera {#init_camera type=manual required=false}

Set up your IP camera and grab its RTSP URL.

1. Connect the IP camera and the Jetson to the same network (or PoE switch)
2. Find the camera's IP — check your router's DHCP list, or use the camera vendor's discovery tool
3. Log into the camera's web UI, make sure RTSP streaming is enabled (it usually is by default)
4. Note down the RTSP URL — you'll paste it in Step 2

Common RTSP URL formats:

• Hikvision: rtsp://admin:password@<ip>:554/Streaming/Channels/101 (main) or /102 (sub)
• Dahua: rtsp://admin:password@<ip>:554/cam/realmonitor?channel=1&subtype=0 (main) or &subtype=1 (sub)
• Generic ONVIF: Use the vendor's ONVIF tool to discover the stream URL

Tip: Test the URL in VLC first (Media → Open Network Stream → paste URL). If VLC can't play it, the deployment won't work either.

Troubleshooting

Step 2: Deploy Industrial Security Service {#deploy type=docker_deploy required=true config=devices/docker_remote.yaml}

Deploy the detection service to your Jetson. The deployer pulls the prebuilt image, writes your RTSP URL into the config, and starts the container.

Target: Local Deployment {#deploy_local type=local config=devices/docker_local.yaml default=true}

Deploy on this machine directly (only works when SenseCraft Solution app is running on the Jetson itself).

Wiring

1. Confirm Docker + NVIDIA runtime are installed on this Jetson
2. Paste the RTSP URL from Step 1
3. Click Deploy

Note: First start takes 1-2 minutes to compile a Jetson GPU-optimized detection engine. The compiled engine is cached, subsequent restarts are instant.

Deployment Complete

The dashboard is at http://localhost:8080. You should see:

• Live annotated video on the main panel
• A yellow restricted zone and a magenta fence line drawn on the frame (defaults — edit them in the browser)
• Stats (FPS, detection count) updating in real time

Quick API check:

curl http://localhost:8080/api/stats
# Expected: JSON with current FPS, detection count, recent events

curl http://localhost:8080/api/events
# Expected: JSON list of triggered events (empty on first deploy)

Troubleshooting

Target: Remote Deployment {#deploy_remote type=remote config=devices/docker_remote.yaml}

Deploy to a remote Jetson over SSH.

Wiring

1. Make sure the Jetson is on the network and reachable via SSH
2. Enter the Jetson IP, SSH username, and password
   • reComputer Industrial defaults: user nvidia / password nvidia (factory) — change if you've already updated
3. Paste the RTSP URL from Step 1
4. Click Deploy — the system will pull the image and start the service

Note: First start takes 1-2 minutes to compile a Jetson GPU-optimized detection engine. The compiled engine is cached, subsequent restarts are instant.

Deployment Complete

The dashboard is at http://<jetson-ip>:8080. You should see:

• Live annotated video on the main panel
• A yellow restricted zone and a magenta fence line drawn on the frame (defaults — edit them in the browser)
• Stats (FPS, detection count) updating in real time

Quick API check:

curl http://<jetson-ip>:8080/api/stats
# Expected: JSON with current FPS, detection count, recent events

curl http://<jetson-ip>:8080/api/events
# Expected: JSON list of triggered events (empty on first deploy)

Troubleshooting

Step 3: Verify in Browser {#dashboard type=manual required=true}

Open the dashboard and tune the safety rules to fit your scene.

1. Open http://<jetson-ip>:8080 in any modern browser on the same network
2. You'll see the annotated live feed with default zones / lines
3. Use the configuration panel (right side) to:
   • Redraw the restricted zone by clicking points on the frame
   • Move the fence line to where you want intrusion alerts
   • Adjust the dwell-time threshold (how long someone has to stand in the zone before it counts)
   • Tune the detection confidence (lower = more sensitive, higher = fewer false positives)
4. Save — changes take effect immediately

Troubleshooting

Deployment Complete

Congratulations! Your industrial security monitoring system is running on the edge.

Quick Verification

1. Walk into the restricted zone in front of the camera — within ~1 second a bounding box should turn red and an event should appear in the event log
2. Cross the fence line — a line-cross event should fire
3. Stand still in the zone past the dwell threshold — a dwell event should fire
4. Hit http://<jetson-ip>:8080/api/events in your browser — the events should be listed there

API Reference

Next Steps

• Hook the /api/events endpoint into your existing alarm / SCADA / WeChat-Work bot
• Pipe events to a central log aggregator (Loki / ELK) for multi-site monitoring
• For multi-camera sites, deploy one instance per Jetson — each one is independent
• Industrial Security Demo on GitHub