A Raspberry Pi 5 dashcam running on QNX 8 that detects vehicles, pedestrians, and traffic lights in real time and warns the rider of danger. It uses the QNX Sensor Framework for camera frames, TensorFlow Lite (SSD MobileNet) for detection, and OpenCV for image work and HSV traffic-light colour analysis. Optional OLED, buzzer, telemetry uploads, and multi-camera are all env-gated.
Prototype only — not a certified driver-assistance or safety system.
This is a monorepo:
├── ai-camera-app/ the QNX C++ application (built + deployed to the Pi)
├── scripts/ bootstrap, build, and deploy tooling
├── presentation/ demo slide deck
└── SETUP.md full environment setup (licensing, the why behind each step)
Everything is cross-compiled on a Fedora host and copied to the Pi to run.
0. Install the QNX SDP (one manual prerequisite — it's proprietary and
license-gated). Install SDP 8.0 into ~/qnx800 via QNX Software Center; see
SETUP.md §2–3. Confirm ~/qnx800/qnxsdp-env.sh exists.
1. Bootstrap the build environment — one command:
git clone https://github.com/akramboussanni/cuRiding-pi.git ~/qnx_workspace
cd ~/qnx_workspace
./scripts/bootstrap.shThis sets up the host cross-build environment — host packages, the
upstream port sources, the Python venv, and the QNX port fixes. It does not
touch the SDP or build the app. Idempotent; SKIP_DNF=1 on non-Fedora hosts.
2. Build the libraries and the app:
source ~/qnx_workspace/env/bin/activate
source ~/qnx800/qnxsdp-env.sh
JOBS=4 ./scripts/build_all.sh3. Deploy to the Pi:
TARGET_HOST=<pi-ip> ./scripts/setup-pi.sh # first time on a fresh Pi
TARGET_HOST=<pi-ip> ./scripts/deploy.sh # subsequent app updatesThen on the Pi: sudo /data/home/qnxuser/qnx-traffic-ai/run.sh.
| Script | Runs on | Purpose |
|---|---|---|
bootstrap.sh |
Fedora host | Stands up the cross-build environment (packages, port sources, venv, fixes) — not the app |
build_all.sh |
Fedora host | Cross-compiles TFLite, OpenCV, and the app |
setup-pi.sh |
→ the Pi | First-time Pi bring-up (camera fix + deploy + launch scripts) |
deploy.sh |
→ the Pi | Ships the app binary + libraries for updates |
Full setup rationale, the Python 3.11/3.13 and CMake notes, and troubleshooting live in SETUP.md. Upstream: the app derives from QNX's ai-camera-app (Apache-2.0 — preserve its headers).