Kolkata's Smart City Challenges: Understanding Sensor Failures During Durga Puja
During the Durga Puja celebrations in Kolkata, a tragic stampede occurred due to sensor failures, revealing critical vulnerabilities in urban technology systems. The reliance on AI-driven video analytics and the absence of LiDAR sensors contributed to the crisis.
Kolkata's Durga Puja Tragedy and Sensor Failures
Did you know that during Kolkata's recent Durga Puja celebrations, a tragic stampede occurred, leading to the loss of 19 lives and injuring 25 others? This incident, which took place at the Chorbagan Sarbojonin Durga Puja pandal, highlighted significant vulnerabilities in urban technology systems. The event, a first in Kolkata's history, has sparked widespread concern across West Bengal. The blame has been directed towards both the police and event organizers for mismanagement, yet the underlying technology, which includes AI-driven video analytics, remains inherently susceptible to structural, weather, and crowd movement failures.
To manage crowd density, Kolkata Police relies on an Integrated Command and Control Centre that processes visual inputs from thousands of CCTV feeds. This system employs AI algorithms to monitor real-time crowd densities and generate alerts when safety thresholds are breached. However, the absence of LiDAR sensors, which are used in other major Indian cities for real-time crowd density estimation, poses a significant challenge. The AI-driven video analytics system, although advanced, struggles during high-density, seasonal festival surges due to specific algorithmic and hardware constraints.
These constraints become evident when we delve into the technicalities. Standard algorithms for crowd-counting rely on Convolutional Neural Networks (CNNs) trained to identify human heads and shoulders. However, when crowd density exceeds four to five people per square meter, individual visual boundaries blur, particularly in heavy rain. This is exacerbated by umbrellas, rain ponchos, and reflections off wet pavement, reducing head-detection accuracy by over 65%. Multi-camera tracking systems face similar challenges, with high crowd density causing shadow clutter and low-light conditions that lead to re-identification errors and misinterpretation of directional flow vectors.
The implications of these technological failures are significant for the industry. The reliance on AI-driven video analytics without the support of LiDAR sensors exposes a critical gap in crowd management during large-scale events. Sayak Basu, a digital journalist, highlights that temporary lightweight structures like pandals are particularly vulnerable to dynamic wet loading, which can double the structural dead-weight of non-waterproof fabric within minutes. This vulnerability necessitates a reevaluation of current systems and the adoption of more robust technologies to prevent future tragedies.
Where Robles.AI Fits
At Robles.AI, we recognize the importance of robust and reliable urban technology systems to manage large-scale events safely. Our advanced AI-driven solutions are designed to enhance real-time crowd monitoring and density estimation, even under challenging conditions such as high-density gatherings and adverse weather. By integrating LiDAR sensors with our AI platforms, we can offer comprehensive solutions that mitigate the limitations experienced by current systems, ensuring enhanced safety and management during critical events like the Durga Puja celebrations.