The Georgia Department of Transportation put Currux Vision through the same four-part evaluation it uses to qualify any video detection system for production use on its signals: missed calls, false positive calls, dropped calls, and count accuracy. The results cleared Currux Vision for video detection statewide — and they were generated with a single camera covering the entire intersection, not a camera per lane or per approach.

That last point matters as much as the accuracy numbers themselves. Most video detection deployments still assume one sensor per movement. GDOT's test intersection ran three actuated movements — Phase 8: Right, Phase 8: Thru, and Phase 3: Left — off one camera, with Currux Vision's AI handling detection, classification, and call logic for the whole approach from a single video feed.

The evaluation ran for 24 continuous hours on September 2, 2025, logging 6,491 vehicle actuations across all three movements. Observed counts were established from time-stamped video review and checked hour by hour against the calls and counts the Currux Vision system generated in the field — the standard protocol agencies use to sign off on a detection system before it goes into production.

Missed calls and dropped calls are the two failure modes that matter most to a signal engineer — a missed or dropped call means a vehicle that should have gotten a green never triggers one. Across all three movements and all 24 hours, Currux Vision missed 3 calls out of 6,491 and dropped none. False positives, which waste green time by calling phases nobody is waiting on, occurred once in 6,491 actuations. Count accuracy against the manually observed ground truth came in at 96.38% — the reported total ran a few percent above the hand-counted baseline, a small, consistent over-count rather than a directional bias toward missing vehicles, and well inside the tolerance GDOT's protocol allows.

Following this evaluation, the Georgia Department of Transportation approved Currux Vision for video-based vehicle detection use across the state — clearing the way for GDOT-governed agencies to specify Currux Vision on new and upgraded signals without a separate case-by-case approval. Paired with the single-camera result, it's a direct answer to the two questions every agency asks before a detection upgrade: does it hold up under independent testing, and does it need to be that expensive to deploy.

Currux Vision designs, builds, and supports its AI edge platform in Houston, TX, serving 300+ agencies and more than 2,000 intersections across the U.S. and Canada. To see single-camera detection validated against your own intersection, request a briefing.