Add YOLOv8n real-time road hazard detection for Jetson Nano

[Copilot]

Bootstraps the project from a README-only state into a functional YOLOv8n inference pipeline targeting the Jetson Nano, with support for CSI/USB cameras, custom model fine-tuning, and optional video/CSV output.

New files

• detect.py — Main inference loop: loads YOLOv8n, reads camera frames, runs model.predict(), overlays bounding boxes, optionally saves annotated video and logs detections to CSV. CLI flags: --config, --weights, --source, --conf, --no-display
• train.py — Fine-tuning wrapper around model.train() with flags for --data, --epochs, --batch, --imgsz, --device, --resume
• utils.py — GStreamer CSI pipeline builder, open_video_source (CSI/USB/file/RTSP), draw_detections, setup_csv_logger/log_detections
• config.yaml — All runtime settings: model weights/confidence/IOU, camera source & resolution, per-class colours, video/CSV output paths
• requirements.txt — ultralytics, opencv-python, numpy, PyYAML, Pillow
• .gitignore — Excludes __pycache__, venv, YOLO run artefacts (runs/, *.pt), output files

Default hazard classes (configurable in config.yaml)

pothole, crack, debris, manhole, bump

Quick start

pip install -r requirements.txt

# Live detection via CSI camera (default)
python detect.py

# USB webcam, custom weights, headless
python detect.py --source 0 --weights runs/train/road_hazard/weights/best.pt --no-display

# Fine-tune on custom dataset
python train.py --data dataset/dataset.yaml --epochs 50 --batch 8

Jetson Nano: install python3-opencv via apt instead of pip to get hardware-accelerated GStreamer support, then drop opencv-python from requirements.txt.

Warning

Firewall rules blocked me from connecting to one or more addresses (expand for details)

I tried to connect to the following addresses, but was blocked by firewall rules:

• dns.google
  • Triggering command: /usr/bin/python python detect.py --help (dns block)
  • Triggering command: /usr/bin/python python train.py --help (dns block)
  • Triggering command: /usr/bin/python python -c import detect; print('detect imports OK') (dns block)
• one.one.one.one
  • Triggering command: /usr/bin/python python detect.py --help (dns block)
  • Triggering command: /usr/bin/python python train.py --help (dns block)
  • Triggering command: /usr/bin/python python -c import detect; print('detect imports OK') (dns block)

If you need me to access, download, or install something from one of these locations, you can either:

• Configure Actions setup steps to set up my environment, which run before the firewall is enabled
• Add the appropriate URLs or hosts to the custom allowlist in this repository's Copilot coding agent settings (admins only)

Original prompt

I have a python based Jetson nano project using yolov8n for computer vision

---

🔒 GitHub Advanced Security automatically protects Copilot coding agent pull requests. You can protect all pull requests by enabling Advanced Security for your repositories. Learn more about Advanced Security.