When Crowd Monitoring Fails Under Pressure

A deadly crush at Kolkata's Chorbagan Sarbojonin Durga Puja pandal, reportedly killing 19 people and injuring 25, has raised serious questions about crowd management, temporary event infrastructure and the limits of surveillance technology during extreme weather. The incident has renewed scrutiny of how police, organisers and city authorities prepare for sudden surges of visitors when narrow streets, heavy rain and dense crowds create dangerous conditions.

The immediate questions concern crowd control and responsibility. But the wider issue goes beyond whether officials were watching CCTV screens or whether organisers deployed enough personnel. Technology can identify warning signs, yet it cannot guarantee that a crowded passage will remain safe when people lose room to move, temporary structures face pressure and emergency routes become obstructed.

Kolkata's experience highlights a problem facing cities across India during large religious festivals. Digital surveillance can improve visibility, but its effectiveness depends on camera placement, image quality, alert systems and the speed at which people respond to warnings.

Why AI Cameras Struggle in Dense, Rain-Soaked Crowds

Kolkata Police relies on an Integrated Command and Control Centre that receives visual information from thousands of CCTV cameras across the city. AI-driven video analytics and computer vision can help identify congestion and support crowd management. However, these systems do not eliminate the physical limitations of surveillance. The police force does not use LiDAR sensors for this purpose, according to the information provided for this report.

The first problem is visibility.

Many crowd-counting systems use computer vision models to estimate the number and distribution of people in an image. As crowds become denser, individuals overlap. Umbrellas conceal heads and shoulders, while rain, reflections, poor lighting and camera angles make identification harder. Research published at the 2023 International Conference on Computer Vision found that crowd-counting models can struggle under adverse weather conditions

The second problem is movement tracking. Systems that follow people across multiple cameras can struggle when visibility deteriorates or pedestrians become difficult to distinguish. Umbrellas and changing light conditions introduce additional errors. Research on pedestrian identification in rainy conditions has documented these problems.

A camera may still produce an image. That does not mean the software can accurately interpret it.

The distinction matters during a crowd surge. A system might detect congestion but fail to estimate how quickly a bottleneck is worsening. An alert may arrive only after people have already lost the space needed to move safely.

Claims that accuracy necessarily falls by a specific percentage in heavy rain should be treated cautiously unless supported by tests of the actual system deployed at the venue. The performance of any model depends on its design, training data, camera position and operating conditions.

Temporary Pandals Face a Different Set of Risks

Technology is only one part of the safety equation. Temporary festival structures introduce engineering risks that require separate assessment.

Pandal construction can involve fabric coverings, metal frames, support columns, barricades and narrow entry and exit routes. These components must withstand the conditions expected at the site, including rain, wind and pressure from surrounding crowds.

Waterlogged fabric can add weight to a structure. Wind can increase lateral loads, while a dense crowd pressing against a barrier can introduce forces that the structure was not designed to withstand. Poor drainage can also leave floors slippery and complicate evacuation.

The applicable structural design requirements, including relevant provisions of Indian Standard IS 875, must be assessed according to the specific structure and loading conditions. Compliance cannot be established simply because a temporary installation received permission to operate.

Nor should the precise cause of the Chorbagan disaster be assumed before investigators establish what happened. The available information does not independently establish that a structural failure caused the deaths or that rainwater doubled the weight of a particular structure. Those possibilities require site inspections, engineering evidence and a verified reconstruction of events.

The immediate investigation must establish the sequence of crowd movement, the condition of access routes, the adequacy of barriers and the response to the sudden rush. Police and organisers also need to account for how warnings were communicated and whether entry controls remained effective as conditions changed.

Who Is Responsible When Safety Systems Fail?

The tragedy raises questions that extend beyond equipment procurement. Authorities must examine whether crowd limits were appropriate, whether trained personnel were available and whether emergency exits remained accessible throughout the event.

Organisers need clear procedures for suspending entry when congestion reaches unsafe levels. Police and control-room operators need authority to act on warnings rather than simply monitor them. Weather alerts should trigger operational changes before conditions deteriorate, including restrictions on entry and coordinated evacuation where necessary.

Insurance and liability arrangements also deserve scrutiny. Investigators must determine whether the relevant parties fulfilled their legal duties, whether safety conditions were breached and whether any failures contributed to the deaths. Responsibility cannot be assigned solely on the basis of who owned a camera, installed a barrier or issued an event permit.

The West Bengal government has announced a probe into the incident and safety checks at other festival venues. Those measures will matter only if they lead to enforceable standards and independent verification.

The Lesson Beyond Kolkata

The Chorbagan disaster exposes the limits of treating smart city technology as a substitute for sound safety planning.

AI can help officials identify crowding, but it cannot make a narrow exit wider, strengthen a weak support column or physically separate people moving in opposite directions. Sensors are useful only when their warnings lead to timely decisions.

The priority now is to establish exactly how the crush unfolded, identify preventable failures and ensure that future festival plans account for heavy rain, sudden crowd surges and the structural demands placed on temporary venues.