A Scientific Reports paper published October 18 evaluates a Gold-YOLO-v8 variant for detecting eye closure, yawning, head turns and phone use in driver video. The authors report mAP@0.5 of 0.97598 versus 0.97345 for the YOLO-v8n baseline, an absolute gain of 0.00253. The study used a public video dataset; the authors say multi-vehicle field validation is still future work.
GLOBAL IMAGING BRIEF
Analysis: The result gives in-cabin vision teams another model configuration to compare, but the small benchmark gain alone is not a reason to change camera hardware or infer safety performance in production vehicles.
Original source
Original peer-reviewed paper published October 18, 2026. Results are limited to the study's public video dataset.