Gujarat Has 80,000 CCTV Cameras. The Hard Part Is Getting Them to Talk

Gujarat, a state of roughly 70 million people on India's western coast, has spent more than a decade stitching together a public surveillance network one contract at a time. The result is more than 80,000 government CCTV cameras, deployed by different police units, municipal bodies, and smart-city programs, each one running on its own vendor's hardware and video-management software. The Gujarat Police Innovation Challenge 2026, expected to start in September, is built around a single question: whether outside teams can build the software layer that ties all of those cameras into a unified, searchable network without forcing the state to rip out what it already owns.

The contest is being marketed as India's largest hackathon devoted to CCTV integration and AI-based video analytics, with ₹37 lakh, roughly $44,000, in total cash prizes across the competition. The technical problem it is testing is older and harder than the prize pool suggests. When a state buys cameras in waves, with 5,000 from one tender and 10,000 from another, the footage ends up scattered across different software stacks, file formats, and storage systems. An investigator looking for a specific vehicle today may be able to pull a clip from a highway camera, but the same plate will not show up when they query a municipal camera two blocks away, because the two systems do not share a database.

What the hackathon is asking contestants to build, in effect, is a translation and federation layer: middleware that sits on top of the existing patchwork and lets a single query reach across vendor boundaries. The capabilities the organizers want to test include automatic number-plate recognition, vehicle tracking across cameras, watchlist matching, and cross-camera search, according to the Open Magazine feature that first reported the contest. The technical partners for the challenge are i-Hub Gujarat, with knowledge support from DA-IICT and the National Forensic Sciences University.

That kind of integration is not a research problem in the way a new AI model is a research problem. The standard for video-management software, ONVIF, has existed for years, and most modern cameras can stream to a conformant system. The harder question is what happens when one city uses one vendor's management platform, a neighboring city uses another, and neither has an upgrade path that includes sharing data with the other. Public-sector CCTV procurement rarely forces interoperability as a condition of purchase, so the fragmentation compounds over time. The wire coverage of the announcement, including Prokerala and ANI, has largely echoed the contest's framing rather than probed the procurement pattern behind it.

The hackathon model has obvious appeal. It draws in students, startups, and established companies, costs the state a fraction of what a custom integration contract would, and gives Gujarat Police a way to compare competing approaches in parallel under the same judging rubric. The same model has obvious limits. A working demo on a 50-camera testbed is not the same thing as a deployment across 80,000 cameras and dozens of operating departments, each with its own procurement office, storage retention rules, and political oversight. Vendor lock-in also cuts both ways. Asking a winning team to build a layer on top of an existing vendor's stack implicitly extends that vendor's position, and the procurement decision that follows the contest will shape which platform emerges as the de facto state standard.

If the procurement rules behind the fragmentation are not changed, September's contest will produce a polished demo at best. Other Indian states that bought cameras in similar waves are about to hit the same wall.