name: no-code-edge-ai description: Deploy the No-Code Edge AI Tool (Node-RED based) on reComputer Jetson Nano for drag-and-drop object detection using a live camera. Requires Jetson Nano with JetPack 4.6.1 and a V4L2 USB camera.
No-Code Edge AI Tool on Jetson Nano
Execution model
Run one phase at a time. After each phase:
- Relay all command output to the user.
- If output contains
[STOP]→ stop immediately, consult the failure decision tree below. - If output ends with
[OK]→ tell the user "Phase N complete" and proceed to the next phase.
Prerequisites
| Requirement | Minimum |
|---|---|
| Hardware | reComputer J1010/J1020 with Jetson Nano module |
| Camera | Logitech C270 HD or other V4L2 USB camera supported by Jetson |
| JetPack | 4.6.1 (R32.7.1) |
| Display | Monitor + keyboard/mouse (or SSH/VNC) |
| Internet | Required for initial Docker download |
Note: Only Jetson Nano is supported. Xavier NX is not supported at this time.
Phase 1 — Preflight
Verify JetPack version and camera connection.
cat /etc/nv_tegra_release
ls /dev/video*
Expected: R32.7.1 (JetPack 4.6.1), at least one /dev/videoN device. [OK] when both pass. [STOP] if wrong JetPack version or no camera detected.
Phase 2 — Download and deploy Docker environment
git clone https://github.com/Seeed-Studio/node-red-contrib-ml.git
cd node-red-contrib-ml
sudo ./docker-ubuntu.sh
The installation and startup takes approximately 7–9 minutes.
Verify Docker containers are running:
sudo docker image ls
sudo docker ps
Expected: three Docker images listed, containers running. [OK] when all containers are up. [STOP] if any container is missing.
Phase 3 — Configure and run object detection (manual)
- Open Chrome browser on the Jetson and navigate to
http://127.0.0.1:1880. - In the Block Area, find the "seeed recomputer" section with three blocks:
- video input — camera source selection
- detection — model selection (COCO dataset)
- video view — output display
- Drag all three blocks to the Programming Area and connect them left-to-right:
video input → detection → video view. - Double-click video input → select Device type (local camera), choose your camera, set resolution.
- Double-click detection → select Model name (COCO dataset).
- Click Deploy in the top-right corner.
[OK] when the video stream shows detected objects with bounding boxes and confidence values.
Phase 4 — Troubleshooting Docker (if needed)
If Docker did not start successfully or "seeed recomputer" blocks are missing:
cd node-red-contrib-ml/
sudo docker-compose --file docker-compose.yaml down
sudo docker-compose --file docker-compose.yaml up
If results are not showing or debug errors appear:
sudo docker image ls
# Verify all 3 required images are present
sudo docker ps
# Verify containers are running
[OK] when blocks appear and detection works after restart.
Failure decision tree
| Symptom | Action |
|---|---|
cat /etc/nv_tegra_release shows wrong version | Reflash with JetPack 4.6.1. This tool only supports R32.7.1. |
No /dev/video* devices | Check USB camera connection. Try a different USB port. Verify camera is V4L2 compatible. |
docker-ubuntu.sh fails | Check internet connectivity. Ensure Docker is installed: sudo apt install docker.io. |
| Missing Docker images after install | Re-run sudo ./docker-ubuntu.sh from the node-red-contrib-ml directory. |
| Node-RED UI not loading at port 1880 | Restart Docker containers with docker-compose down then up. Check sudo docker ps. |
| "seeed recomputer" blocks missing | Docker containers not fully started. Restart and wait for all 3 containers to be running. |
| Wrong resolution causes runtime error | Double-click video input block and select the correct resolution for your camera. |
Reference files
references/source.body.md— full original Seeed tutorial with block diagrams, UI screenshots, email notification project example, and advanced block operations (reference only)