THIRUVANANTHAPURAM: The state police have onboarded an internally developed AI-based tool to help personnel identify objects from visual data. The tool — developed by Thrissur city police in January this year — was selected by the department for statewide use, assessing that it can help tremendously in criminal investigation.
Drawn up by ASP Hardik Meena, an IIT graduate now serving with the Perumbavoor sub-division, the tool can detect vehicles and pedestrians from any visual including CCTV footage.
What makes the project stand out is that the tool was fully developed in-house without any cost to the government. Two other technically-qualified officers -- Thrissur Range DIG Harishankar and Thrissur city police commissioner S Ilango -- monitored the process.
A senior officer told TNIE that the tool was initially developed to identify motorcycles but has now been augmented to identify four-wheelers, trucks and pedestrians.
“Work is now in progress to make the tool identify various pedestrian subsets,” the officer said.
Significantly, the tool will trim the time required to identify various elements from a visual.
“For instance, if a murder-accused used bikes, then the cops need not sit for hours scrutinising the entire CCTV footage from the location. Instead, they can command the tool to identify bikes. All frames with bikes will be available within a matter of minutes,” the officer said.
With AI-based cameras in short supply, the police department can, for the time being, rely on the tool which can be embedded with existing cameras to access some smart features.
Police sources said coding experts with the force are now layering the codes to augment the effectiveness of the tool. It has been programmed to work even if the visuals are grainy and could help investigators get the best frame from the footage.
In its advanced form, the tool could detect people too from visual data.
“The tool is now being trained to identify pictures. That’s done by providing it access to images on open sources,” an officer said.