Lab notes: computer vision in practice
Short write-ups of what we've built and learned — real footage, real numbers, and what actually made the difference.
Making a tank disappear: removing moving objects from real video
How we removed a moving tank from real, panning video by rebuilding the background from other frames, not capturing an empty one.
Read → Road safetySeeing cones is not the same as understanding a lane closure
Why detecting traffic cones is easy, and why understanding the lane closure they create is the part that matters for road safety.
Read → Queue analyticsWhy airport queue counts run low — and how a top-down camera fixes it
Under angled cameras 67% of tracked people overlap. A top-down view and the right model found about 300 people per frame.
Read → Drones & geospatialFrom pixels to latitude and longitude: object detection you can put on a map
Vision models find objects in aerial images easily. Turning them into real coordinates and real sizes is what makes the result useful.
Read → Emergency responseDetecting fire, smoke and vehicles in one pass
No off-the-shelf model detects cars, trucks, fire and smoke together. Running several models per frame did — and resolution mattered most.
Read → Workplace safetyReal-time PPE tracking without buying new cameras
Turn the CCTV you already have into a hard hat and hi-vis vest checker that sends verified alerts and a daily compliance report.
Read →Have footage like this?
Book a free 15-minute call, or email us a sample image or short clip. We'll tell you honestly whether computer vision can do the job.