Validated Adaptation for Aerial Crowd Monitoring at Mass Gathering Scale: A Deployment Protocol, a Severity Law, and a Diagnostic for Label-Free Drone Crowd Counting, Toward the FIFA World Cup 2034 (Saudi Arabia)
Researchers have developed a method to accurately count crowds using drones without labeled training data. This is crucial for large events like the FIFA World Cup in Saudi Arabia in 2034. The team tested their approach on footage from Hajj gatherings and found it can recover up to 49% of errors caused by changes in crowd density. They also established a 'severity law' to determine when a method's performance degrades, and a 'stability budget' to identify safe configurations
Researchers have developed a method to accurately count crowds using drones without labeled training data. This is crucial for large events like the FIFA World Cup in Saudi Arabia in 2034. The team tested their approach on footage from Hajj gatherings and found it can recover up to 49% of errors caused by changes in crowd density. They also established a 'severity law' to determine when a method's performance degrades, and a 'stability budget' to identify safe configurations for drone deployment. This work aims to improve crowd management at large events.
---
Why it matters: This matters because accurate crowd counting can help prevent crushes and save lives in densely populated areas. The ability to adapt to changing crowd dynamics without labeled data is a significant breakthrough in AI research, with potential applications beyond sports events.
Source: https://arxiv.org/abs/2608.17625
This article was originally published at: https://arxiv.org/abs/2608.17625