AI / computer vision · Python + YOLOv8 · Personal project
R.A.S — it watches your ride footage the way a spotter would.
A desktop tool for moto vloggers: feed it raw POV footage, and it detects vehicles, pedestrians, and road hazards frame-by-frame, overlaying threat-level markers with a cinematic HUD look — while keeping your original audio untouched.
The problem
Raw POV riding footage is honest. It's also flat.
A GoPro strapped to a helmet captures everything, but it doesn't help a viewer feel the near-misses, the traffic reads, the moments that actually made the ride tense. Editing that in by hand, shot by shot, doesn't scale past one video. R.A.S exists to do it automatically, frame by frame, across an entire ride.
What it does
Object detection, turned into a story a viewer can feel.
Frame-by-frame detection
YOLOv8 identifies vehicles, pedestrians, and road elements in every frame — not a sampled subset, the whole ride.
Threat-level classification
Detected objects are scored Red, Yellow, or White by how much of the frame they occupy — a rough proxy for proximity and urgency.
GPU-accelerated rendering
CUDA auto-detected, live FPS telemetry during render, 4K support — built to actually finish a real ride video, not just a demo clip.
Audio untouched
Original engine, wind, and commentary audio is preserved and muxed back into the final export — this is a visual overlay, not a re-edit of the ride.
Why it holds up
Ships as a real .exe, not just a script
Packaged and distributed as a standalone Windows build — no Python install, no FFmpeg setup, double-click and load a video.
Built for real GPUs, not just a benchmark
CUDA detection, live FPS during render, and graceful cancellation — the render pipeline is built to be actually used, not just demoed once.
Honest about what it is
A post-processing content tool, not a real-world collision-detection or safety system — that line is drawn deliberately, in the product itself.
Want the real build story — the detection pipeline and the render architecture?
Read the full case study